UAV-Based Spectral and Thermal Indices in Precision Viticulture: A Review of NDVI, NDRE, SAVI, GNDVI, and CWSI

Abstract

Unmanned aerial vehicles (UAVs) with multispectral sensors are transforming precision viticulture by enabling detailed monitoring of vineyard variability. Vegetation indices such as NDVI, NDRE, GNDVI, and SAVI are widely applied to estimate vine vigor, canopy structure, and water status. Beyond agronomic traits, UAV-derived indices can inform grape composition, including sugar content (°Brix), total phenolics, anthocyanins, titratable acidity, berry weight, and yield variables measurable in the field or laboratory to validate spectral predictions. Strengths of UAV approaches include high spatial resolution, rapid data acquisition, and flexibility across vineyard blocks, while limitations involve index saturation in dense canopies (e.g., Merlot, Cabernet Sauvignon), environmental sensitivity, and calibration requirements across varieties and phenological cycles. Integrating UAV data with ground-based measurements (leaf sampling, yield mapping, proximal or thermal sensors) improves model accuracy and stress detection. Abiotic stresses (water deficit, nutrient deficiency) can be distinguished from biotic factors (pest and fungal infections), supporting timely interventions. Compared to manned aircraft or satellite platforms, UAVs offer cost-effective, high-resolution imagery for precision vineyard management. Future directions include combining UAV indices with machine learning and data fusion to predict grape maturity and wine quality, enhancing decision-making in sustainable viticulture and precision enology.

1. Introduction

Grapevine (Vitis vinifera) is among the most economically crucial perennial crops worldwide, with production exceeding 80 million tons in 2022 [1]. The wine sector remains a major global industry, with worldwide consumption reaching 234 million hectoliters in 2021. France and Spain continue to occupy leading positions in global production and trade, exemplified by their centuries-old appellation systems and regulated denominations of origin, which ensure wine quality and protect regional identity [2]. In 2023, Spain ranked as the second-largest wine exporter by volume, shipping approximately 20.9 million hectoliters, and as the third by export value, contributing more than EUR 23.7 billion annually to its gross value added, representing 2.2% of the national total [3]. Viticulture is practiced in a broad spectrum of climatic regions, including Mediterranean, continental, oceanic, and semi-arid climates, and on a diversity of soil types such as sandy, clay, loam, limestone, and volcanic soils. Despite this adaptability, grapevines remain highly sensitive to biotic stress factors such as fungal diseases like black rot, downy mildew, and botrytis; insect pests including grapevine moths and phylloxera; and viral or bacterial infections like leafroll virus, as well as abiotic stress factors such as drought or waterlogging, extreme temperatures, soil salinity, and nutrient imbalances, all of which can severely compromise their physiology, productivity, and quality potential [4]. Among these challenges, fungal diseases such as black rot and downy mildew are particularly damaging, often exacerbated by excessive humidity (relative humidity levels above 80–90% that create favorable microclimatic conditions for fungal proliferation within the canopy) and inadequate irrigation management, including overwatering practices that maintain constant soil moisture, irregular irrigation scheduling that creates prolonged periods of elevated soil humidity, and failure to adjust water application according to climatic conditions during rainy or humid seasons [5]. These management deficiencies result in humid microclimates with poor air circulation within the canopy, conditions that facilitate fungal infection and development. Vegetation indices derived from UAV-based multispectral imagery, such as NDVI, NDRE, SAVI, and GNDVI, are widely used to assess canopy vigor, structure, and water status, offering valuable information on the physiological condition of vines under such biotic and abiotic stress conditions.

Grapevine phenology plays a critical role in UAV-derived spectral responses throughout the crop cycle. Stages such as budburst, flowering, fruit set, véraison, and harvest maturity exhibit distinct physiological and biochemical changes, including variations in canopy structure, leaf area, chlorophyll concentration, and berry development [6]. These dynamic processes influence spectral reflectance and vegetation indices such as NDVI, NDRE, GNDVI, and SAVI [7]. For instance, limited leaf area and soil background exposure during early growth can reduce index values, whereas dense canopies and high chlorophyll concentrations at véraison may lead to index saturation [8,9,10]. Aligning UAV flight campaigns with key phenological stages allows better discrimination of variability caused by vine growth stage, environmental stress, or management practices, thereby enhancing the predictive power of UAV-based assessments for vigor mapping, water status estimation, and ripening monitoring (Figure 1). Understanding these spectral variations across phenological stages is particularly relevant under current climate change conditions, where shifts in temperature and water availability increasingly affect vine physiology and spectral behavior.

Climate change further reshapes viticultural dynamics, as rising temperatures accelerate phenology, alter berry composition, and increase the frequency of irregular water availability, which in turn amplifies pest pressure from insects such as grapevine moths and spider mites and promotes fungal diseases like powdery mildew and downy mildew [12]. In this context, advanced technologies such as unmanned aerial vehicles (UAVs) equipped with multispectral and thermal sensors are proving invaluable, as they enable high-precision monitoring of canopy health [13], soil moisture, and plant water status. They optimize water resources by enabling targeted irrigation scheduling based on spatial variability in vine water demand, reducing unnecessary water use. UAV data also allow early detection of stress symptoms such as water stress, nutrient deficiencies, or early disease onset through analysis of spectral indices (NDVI, NDRE) and thermal imagery (canopy temperature maps). These technologies enhance sustainability and resilience in viticulture [14]. Given their ability to capture high-resolution spectral and thermal information, UAV-based systems have become central tools in precision agriculture, bridging the gap between environmental monitoring and practical vineyard management.

Precision agriculture technologies, including unmanned aerial vehicles (UAVs), proximal sensors, soil moisture probes, and satellite-based monitoring, have become valuable tools for vineyard management, with UAVs emerging as particularly effective for acquiring multispectral and thermal sensor data over large areas in a non-invasive manner [15]. Their ability to capture high-resolution imagery with spatial resolutions ranging from 1 to 5 cm provides a high-throughput alternative to traditional monitoring methods, including manual leaf sampling, visual crop scouting, ground-based phenological assessments, conventional meteorological station measurements, and laboratory-based plant tissue analysis, enabling detailed assessments of canopy vigor, disease incidence, pest infestations, and vine water status [16,17,18,19,20,21]. Compared to satellite platforms commonly used in precision viticulture, such as Landsat 8/9 (30 m native resolution, pan-sharpened to 15 m) and Sentinel-2 (10–20 m), which typically offer spatial resolutions of 5–10 m per pixel, UAV-mounted sensors can achieve pixel sizes as small as 0.5 cm, thereby enabling monitoring at the level of individual leaves or vines [22].

Beyond their operational advantages, UAV-mounted sensors provide crucial insights into vine physiology by capturing reflectance across multiple spectral domains: visible (400–700 nm), near-infrared (NIR, 700–1300 nm), red-edge (700–740 nm), shortwave infrared (SWIR, 1300–2500 nm), and thermal infrared (>8000 nm) each linked to specific physiological and biochemical traits [10]. For example, reflectance in the red region around 660 nm is strongly absorbed by chlorophyll, while the NIR region beyond 700 nm shows a sharp increase due to internal leaf scattering, both key indicators of photosynthetic capacity and canopy structure [23,24]. These spectral properties are further influenced by pigment composition (chlorophyll a/b and carotenoids), leaf anatomical structure (mesophyll and intercellular air spaces), and soil background characteristics (moisture, organic matter, texture, and surface roughness), as well as understory vegetation, shadows, and crop residues.

The spectral behavior of vegetation typically exhibits low reflectance in the visible region, strong absorption in the red band, and a sharp rise in the NIR, producing distinct signatures that reflect the plant’s physiological condition. Figure 2 illustrates these characteristic reflectance patterns of healthy and stressed vegetation, highlighting their diagnostic value for assessing canopy health. Understanding these spectral responses is essential for designing and interpreting vegetation indices and requires careful consideration of soil reflectance variability to ensure accurate assessments of vine condition. According to the literature, the water content of a healthy leaf typically ranges between 70 and 80%, whereas stressed leaves may drop to 50–60%. The spectral profiles in Figure 2 correspond to single-layer canopies under normal conditions and represent general vegetation behavior without species specificity [24]

Building on these spectral foundations, vegetation indices derived from UAV imagery provide quantitative proxies for key vine traits by capturing physiological and structural information. Indices based on visible wavelengths primarily reflect chlorophyll and carotenoid concentrations, indicating photosynthetic activity and stress responses [25,26,27]. Near-infrared indices are closely linked to canopy architecture, leaf area index (LAI), and vine vigor, while water-related and thermal indices detect variations in transpiration, stomatal conductance, and leaf temperature [28]. Together, these parameters form the foundation for predictive frameworks that support precision irrigation, stress detection, and yield forecasting in viticulture. Looking ahead, emerging approaches that integrate spectral indices with machine learning and hybrid multispectral–thermal–structural models are expected to enhance predictive accuracy, enable dynamic monitoring across phenological stages, and strengthen the role of UAV-based sensing as a decision-support tool for sustainable and climate-resilient vineyard management [14,29,30].

Ensuring accurate and reliable remote sensing assessments in viticulture requires careful sensor selection, precise spectral filtering, and rigorous radiometric calibration [31]. In practice, vineyard monitoring is commonly performed using commercially available multispectral cameras, such as the MicaSense RedEdge (MicaSense Inc., Seattle, WA, USA) and Parrot Sequoia (Parrot SA, Paris, France), which capture spectral radiance across selected wavelength regions and convert it into reflectance an intrinsic property representing the fraction of incident radiation reflected by the target. Sensor performance depends on spatial, spectral, and radiometric resolution, with spatial resolution determining the ground area represented by each pixel [32,33,34]. UAV-based sensors can resolve fine-scale canopy features such as individual leaves or disease symptoms (e.g., powdery mildew, downy mildew, or leaf necrosis), although this high level of detail may reduce flight coverage depending on focal length, detector size, and flight altitude [35,36]. Spectral resolution, in contrast, defines the number and width of spectral bands. Narrower bands (10–20 nm) offer greater sensitivity to biochemical traits such as chlorophyll content, water status, or disease presence, whereas broader bands capture general canopy information but with lower specificity [28,37,38]. Multispectral cameras typically acquire 3–10 bands with bandwidths ranging from 10 to 100 nm, depending on sensor design and application [39]. For example, red and NIR bands are used to compute NDVI for canopy vigor monitoring, while water stress detection benefits from red-edge bands, and disease surveillance requires high spatial and spectral resolution.

High-resolution multispectral imagery acquired by UAVs enables the calculation of vegetation indices that serve as quantitative proxies for assessing vine vigor, canopy health, and water status. Among these indices, the Normalized Difference Vegetation Index (NDVI) remains the most widely applied due to its sensitivity to chlorophyll content and photosynthetic activity [40]. NDVI has been extensively correlated with chlorophyll concentration and leaf structure, proving effective in identifying intra-vineyard variability related to plant vigor, grape yield, and even berry quality parameters [41]. When combined with complementary agronomic data, such as soil apparent electrical conductivity, canopy volume, or cluster counts, NDVI-based maps provide valuable insights into vineyard heterogeneity, facilitating site-specific management and harvest strategies [42]. Urretavizcaya et al. [43], for instance, demonstrated that NDVI integrated with soil and yield data enabled the differentiation of grape lots at harvest based on distinct physicochemical properties, including berry weight (BW), total soluble solids (TSS), titratable acidity (TA), malic acid (MalA), tartaric acid (TarA), yeast-assimilable nitrogen (YAN), total anthocyanins (TAnt), extractable anthocyanins (EAnt), and total phenolics (TP).

Beyond vigor assessment, NDVI has also proven valuable for detecting vine diseases. UAV-derived NDVI values have been correlated with symptoms of grapevine leaf stripe disease, enabling spatial discrimination between symptomatic and asymptomatic vines [44]. Similarly, NDVI values tend to decrease with increasing severity of powdery mildew in Vitis vinifera ‘Thompson Seedless,’ highlighting its potential as an indirect diagnostic tool for vineyard health surveillance and targeted phytosanitary interventions [45]. More recently, NDVI and related indices such as RDVI (Renormalized Difference Vegetation Index), OSAVI (Optimized Soil-Adjusted Vegetation Index), and NDRE (Normalized Difference Red-Edge Index) have been used to estimate vine water status, showing moderate-to-strong correlations with stem water potential (Ψstem). These correlations further improve when red and red-edge bands are integrated into multivariate regression models [46,47].

Unlike climacteric fruits such as apples or bananas, grapes are non-climacteric and do not continue ripening after harvest. Thus, harvesting at the appropriate phenological stage is crucial to preserving berry quality. Premature or delayed harvests can negatively affect sugar accumulation, acidity balance, phenolic composition, and aromatic profile, all of which determine wine quality through their influence on potential alcohol, freshness, structure, and varietal expression. Traditional methods for evaluating grape maturity based on total soluble solids (TSS), pH, and titratable acidity are destructive, labor-intensive, and spatially limited [48,49]. In recent years, UAV-based multispectral imagery has emerged as a non-destructive alternative for monitoring ripening-related traits. Indices such as NDVI, NDRE, SAVI, and GNDVI have shown strong relationships with biochemical changes in grapes, including variations in carotenoids, soluble solids, and aromatic precursors such as monoterpenes (linalool, geraniol), norisoprenoids (β-damascenone), and methoxypyrazines. These correlations link spectral information to wine composition and quality [22,26,50,51,52]. This capability to monitor physiological and biochemical dynamics non-invasively allows for more precise harvest timing and the development of site-specific enological strategies.

Although numerous reviews have examined UAV applications and vegetation indices in viticulture [7,15,29,53,54,55], most have focused on vigor zoning, irrigation scheduling, or disease detection without systematically linking spectral information to grape or wine composition. For instance, Sassu et al. (2021) [15] provided an overview of UAS technologies and vineyard variability, Giovos et al. (2021) [53] critically assessed vegetation indices, and Tardaguila et al. (2021) [55] summarized digital tools for vineyard management. Similarly, Ammoniaci et al. (2021) [54], Ferro and Catania (2023) [29], and Singh et al. (2022) [7] discussed monitoring technologies and bibliometric trends but did not explore the connection between UAV-derived NDVI, NDRE, SAVI, GNDVI, or CWSI and grape biochemical composition or enological outcomes.

In contrast, studies explicitly linking UAV-derived spectral indices with wine quality attributes, including alcohol content, acidity, color intensity, phenolic composition, tannin structure, aromatic complexity, and overall sensory quality, remain scarce, and this gap has not yet been systematically synthesized. Therefore, the present review aims to integrate and critically evaluate current knowledge on the application of UAV-based vegetation indices, particularly NDVI, NDRE, SAVI, GNDVI, and CWSI for monitoring grapevine water status, vigor, and maturity, with special emphasis on their implications for wine composition and quality. By addressing this underexplored dimension, this work seeks to bridge the gap between technological advances in UAV-based remote sensing and their practical application to improve grape and wine quality within the paradigm of precision and sustainable viticulture.

2. Methodology of Literature Selection

The methodological workflow of this review is summarized in Figure 3, which illustrates the main stages of the literature selection and analysis process. The approach comprised five sequential steps: (i) database search (Scopus, Web of Science, and Google Scholar) covering the period 2015–2024; (ii) preliminary screening of titles and abstracts; (iii) application of inclusion and exclusion criteria; (iv) extraction and categorization of relevant information by spectral index and viticultural application; and (v) synthesis and visualization of trends. This structured workflow ensured methodological transparency, consistency, and reproducibility throughout the review.

This review followed an integrative and systematic approach to analyze scientific literature on the use of unmanned aerial vehicles (UAVs) equipped with multispectral sensors in precision viticulture from 2015 to 2024, emphasizing spectral indices, their correlation with agronomic parameters (e.g., water stress, vine vigor, yield prediction), and their contribution to berry and wine quality. Rather than providing a general overview, the review focuses on applications that directly support vineyard decision-making, including vigor zoning, disease detection, irrigation scheduling, and quality assessment. Data collection was performed through manual searches of three major scientific databases: Scopus (https://www.scopus.com/, accessed on 25 October 2024), Web of Science (https://www.webofscience.com/, accessed on 25 October 2024), and Google Scholar (https://scholar.google.com/, accessed on 25 October 2024), using keyword combinations including “Unmanned Aerial Vehicles (UAV)”, “Precision Viticulture”, “Multispectral Imaging”, and “Vegetation Indices”. No automated bibliometric packages in R (version 4.3.2, R Core Team, Vienna, Austria) or Python (version 3.10, Python Software Foundation, Wilmington, DE, USA) were used; instead, all articles were individually screened and analyzed. The search initially identified 150 articles, and after applying inclusion and exclusion criteria, 80 studies were selected for detailed review.

Inclusion criteria considered peer-reviewed studies on UAV-based spectral analysis in vineyards that reported quantitative validation metrics such as coefficient of determination (R², with significance levels explicitly stated when available), root mean square error (RMSE), mean absolute error (MAE), or statistical significance tests (e.g., ANOVA, Tukey’s HSD, Pearson correlation). Exclusion criteria eliminated studies centered on other crops such as wheat, maize, or rice, as well as reviews without methodological contributions or works lacking analytical rigor.

Extracted information included sensor types (RGB, multispectral, thermal), spectral indices used (e.g., NDVI, NDRE, SAVI, GNDVI, CWSI), and their practical applications, such as water stress detection, vigor mapping, disease monitoring, and site-specific management strategies. The systematic categorization revealed NDVI as the most widely applied index, followed by NDRE, SAVI, and GNDVI. Less frequent indices TCARI (Transformed Chlorophyll Absorption in Reflectance Index) and OSAVI (Optimized Soil-Adjusted Vegetation Index) remain underexplored, likely because they require more complex calibration or are sensitive to specific canopy and soil conditions, making them less common in commercial vineyard operations. These findings are summarized in Figure 4a,b, which present the distribution of indices across measured variables (chlorophyll, disease/pest detection, maturity/quality, water stress, yield) and their frequency of use in viticulture studies.

In addition to the overall frequency of spectral index use, the temporal distribution of studies applying NDVI from 2015 to 2024 reveals notable trends. The number of publications increased steadily from 5 in 2015 to a peak of 14 in 2018, highlighting growing interest in NDVI for monitoring canopy vigor, water stress, and disease detection. Following this peak, a slight decline was observed in 2019–2020, possibly reflecting the exploration of alternative indices such as NDRE and SAVI for specific applications. A resurgence occurred in 2021–2022, with 12 and 16 publications, respectively, after which the number of studies gradually decreased through 2024. These observations suggest that NDVI remains the most widely adopted index in UAV-based viticulture, although its relative novelty has stabilized, with recent research increasingly incorporating complementary indices to address specific physiological or management objectives.

3. UAV Data Acquisition and Processing

The integration of unmanned aerial vehicles (UAVs) into viticulture has transformed how canopy structure, vigor, and stress are monitored. This section outlines the key aspects of UAV data acquisition and image processing that underpin subsequent analyses of spectral and thermal indices.

3.1. General Concepts

The use of drones to obtain detailed data at both individual plant and vineyard-block scales, typically ranging from single vines up to plots of 1–5 hectares, has increasingly been applied in viticulture [56]. Ensuring optimal grape production and consistent wine quality, especially under climate change scenarios, such as increased frequency of heat waves, irregular precipitation, and drought events, requires highly accurate monitoring of vine health and condition, with typical sensor spatial resolutions ranging from 1 to 5 cm per pixel for canopy-level observations [57]. Grapevines are regularly affected by biotic and abiotic stressors including pests, diseases, and environmental fluctuations that alter canopy physiology and chemistry.

UAVs offer higher resolution and greater operational flexibility ground-based methods, such as manual canopy measurements, leaf sampling, destructive biochemical assays, or proximal sensors mounted on tractors [58]. Equipped with a wide range of sensors RGB, multispectral, hyperspectral, thermal, and LiDAR, UAVs can be adapted to specific monitoring objectives, such as estimating canopy vigor, detecting disease outbreaks, assessing water stress, or mapping nutrient variability [59]. Among these, multispectral imagery has emerged as a cost-effective tool in vineyard management, capturing plant reflectance across visible, near-infrared (NIR), and thermal domains [60].

The NIR band (700–800 nm) is particularly important for assessing chlorophyll concentration and canopy development [61]. Because chlorophyll and water strongly absorb radiation in the red and NIR regions, variations in reflectance allow the early detection of physiological imbalances, such as nitrogen deficiency, water stress, or early signs of disease. Chlorophyll content is also closely related to nitrogen availability and thus acts as a proxy for photosynthetic efficiency and nutrient status [25]. Beyond physiology, UAVs combined with LiDAR systems can generate high-resolution 3D maps of canopy geometry, including traits such as leaf area and spatial distribution [62]. Spectral and structural information derived from UAVs can then be used to calculate vegetation indices (e.g., NDVI, GNDVI, PRI) and spatial models (e.g., kriging, regression-based yield prediction), facilitating site-specific decision-making. For instance, zones showing low NDVI values or irregular canopy structure can be targeted for differential fertilization, irrigation adjustments, or selective pruning to optimize vineyard performance.

3.2. Data Processing

Processing UAV imagery involves several stages depending on the specific objectives and applications. Drone flight characteristics, including flight altitude, speed, front and side overlap, and camera exposure settings, are essential to ensure high-quality data acquisition. In vineyard studies, a flight altitude of approximately 30–50 m above the canopy and a drone speed of 3–5 m/s are recommended. Frontlap between consecutive images is typically 75–85%, while sidelap between parallel flight lines is maintained at 65–75% [63]. These settings ensure sharp images, adequate spatial resolution (2–5 cm/pixel), and coherence for 2D mosaicking and 3D reconstructions. Multispectral and thermal cameras are adjusted for gain, shutter speed, and radiometric calibration during image capture.

The technical specifications of UAV sensors commonly reported in viticultural research include multispectral systems such as the MicaSense RedEdge-M (MicaSense Inc., Seattle, WA, USA) and Parrot Sequoia (Parrot SA, Paris, France), which capture five narrow spectral bands (blue, green, red, red-edge, and near-infrared), as well as RGB cameras such as the DJI Phantom 4 Pro (SZ DJI Technology Co. Ltd., Shenzhen, China) with 20 MP resolution and a 94° field of view. These configurations provide ground sample distances of 1–5 cm per pixel depending on flight altitude and are widely used for canopy reconstruction and vegetation index generation. Image processing and analysis are commonly performed using software such as Pix4Dmapper (version 4.5.6, Pix4D SA, Prilly, Switzerland) [64], Agisoft Metashape (version 1.6.3, Agisoft LLC, St. Petersburg, Russia) [65]. A calibration plate or reflectance panel (Figure 5) was used to standardize radiometric measurements.

For standard analyses, particularly those focused on vegetation indices such as the Normalized Difference Vegetation Index (NDVI), two-dimensional (2D) workflows involve the alignment and mosaicking of overlapping images to generate a continuous representation of the vineyard canopy without reconstructing 3D structures. After alignment, multispectral images are processed into an orthomosaic, enabling precise assessment of vineyard health, vigor, and spatial variability (Figure 5). More advanced applications, such as the estimation of canopy structural traits, require three-dimensional (3D) reconstructions generating Digital Surface Models (DSM), Digital Elevation Models (DEM), point clouds, and canopy models, which support measurements of plant height, width, and volume critical parameters for precision pruning, yield estimation, and trellis system optimization [66].

Figure 5. Adapted multispectral image processing workflow [67]. Steps include image acquisition, radiometric calibration, 2D mosaicking, 3D reconstruction, and index calculation.

Figure 5. Adapted multispectral image processing workflow [67]. Steps include image acquisition, radiometric calibration, 2D mosaicking, 3D reconstruction, and index calculation.

The reliability of spectral indices depends on radiometric calibration using reflectance panels, which correct for variations in solar radiation, sensor sensitivity, and atmospheric conditions during image capture. Angular corrections based on DEMs are applied considering the solar zenith angle, slope, and aspect, enhancing the accuracy of both spectral and structural canopy assessments [32,68].

Another challenge arises from non-vine elements such as inter-row vegetation, shadows, and soil, which complicates accurate segmentation. Various methodologies have been proposed (Figure 6). For example, Pádua et al. [69] developed segmentation approaches exploiting structural differences between vine canopy and inter-row vegetation, while clustering algorithms such as k-means (partitional clustering) and CLARA (Clustering Large Applications) are applied to discriminate vegetation, shadows, and soil based on spectral traits. More advanced techniques, including differential digital models [70], UAV-based LiDAR data [71], and deep learning segmentation architectures such as Segmentation Network (SegNet), U-shaped Convolutional Network (U-Net), and Modified Segmentation Network (ModSegNet) [72], have further improved accuracy and robustness under diverse vineyard conditions. These methodologies have been reported to increase the precision of multispectral analyses by up to 10–15% compared to conventional approaches, enabling reliable vineyard monitoring and informed agronomic decision-making.

4. Vegetation and Thermal Indices in Precision Viticulture

Spectral indices synthesize the unique electromagnetic signatures of vegetation and are essential tools for assessing vine status. In precision agriculture, they have been applied to monitor crops, estimate water stress, delineate management zones, and predict yield [73]. Despite their broad utility, not all indices are equally sensitive or suitable for every situation; therefore, their selection must be aligned with both the application purpose and the spectral capacity of the available sensors [53]. In practical terms, spectral indices provide an efficient means of integrating spatial and temporal information through relatively simple computations, allowing the detection of subtle canopy changes that are not visible to the naked eye. Nevertheless, their reliability depends on several factors, including image resolution, radiometric quality, environmental conditions, and viewing geometry. In vineyard environments, additional challenges arise from soil background and shadow effects, which may introduce unwanted variability in canopy reflectance and compromise index stability [63].

Within the framework of precision viticulture, this review focuses on the indices most relevant to vine vigor, chlorophyll content, and water status. These indices can be classified according to the spectral bands they employ and the physiological parameters they estimate. Indices based on the red and near-infrared (NIR) regions, such as NDVI [74], GNDVI [75], and SAVI [76], are particularly useful for evaluating biomass and canopy vigor [77]. In contrast, indices incorporating the green or red-edge regions, such as NDRE [78], are more sensitive to chlorophyll concentration and nutritional status. Although these indices are widely used, their performance can vary across phenological stages and vineyard conditions, highlighting the need to critically assess their suitability and limitations before applying them to quality-oriented viticulture.

Taken together, these vegetation indices provide complementary perspectives on vine vigor, chlorophyll content, and water status, each with specific strengths and limitations (

Table 1. Comparative summary of vegetation indices applied in precision viticulture.

Index	Formula	Main Application	Limitations in Vineyards
NDVI	${\displaystyle \frac{(\mathrm{N}\mathrm{I}\mathrm{R}-\mathrm{R}\mathrm{E}\mathrm{D})}{(\mathrm{N}\mathrm{I}\mathrm{R}+\mathrm{R}\mathrm{E}\mathrm{D})}}$	Vigor mapping, biomass estimation, zoning for grape/wine quality	Saturates at high biomass; influenced by soil background, shadows, and atmospheric effects [79,80]
GNDVI	${\displaystyle \frac{(\mathrm{N}\mathrm{I}\mathrm{R}-\mathrm{G}\mathrm{R}\mathrm{E}\mathrm{E}\mathrm{N})}{(\mathrm{N}\mathrm{I}\mathrm{R}+\mathrm{G}\mathrm{R}\mathrm{E}\mathrm{E}\mathrm{N})}}$	Chlorophyll estimation, nutrient status, yield prediction	Sensitive to soil background, cloud cover, haze; lower performance before ripening [79,81]
SAVI	${\displaystyle \frac{\left(NIR-RED\right)}{\left(NIR+RED+L\right)}}\ast (1+L)$	Vigor assessment under sparse canopy or early growth stages	Requires calibration of L; less effective in dense canopy conditions [76]
NDRE	${\displaystyle \frac{(\mathrm{N}\mathrm{I}\mathrm{R}-\mathrm{R}\mathrm{E})}{(\mathrm{N}\mathrm{I}\mathrm{R}+\mathrm{R}\mathrm{E})}}$	Chlorophyll content monitoring, stress detection, water status	Less robust under heterogeneous soils; indirect indicator; needs calibration with ground data [82,83]
CWSI	${\displaystyle \frac{\left[\right(\mathrm{T}\mathrm{c}-\mathrm{T}\mathrm{a})-(\mathrm{T}\mathrm{w}\mathrm{e}\mathrm{t}-\mathrm{T}\mathrm{a}\left)\right]}{\left[\right(\mathrm{T}\mathrm{d}\mathrm{r}\mathrm{y}-\mathrm{T}\mathrm{a})-(\mathrm{T}\mathrm{w}\mathrm{e}\mathrm{t}-\mathrm{T}\mathrm{a}\left)\right]}}$	Direct indicator of vine water stress, stomatal conductance, irrigation scheduling	Requires wet/dry reference calibration; affected by canopy architecture, time of acquisition [51,84]
LAI	${\displaystyle \frac{\mathrm{L}\mathrm{e}\mathrm{a}\mathrm{f} \mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}}{\mathrm{G}\mathrm{r}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{d} \mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}}}$	Canopy architecture, vigor, photosynthetic capacity, yield estimation	Sensitive to lighting conditions, soil interference, mixed vegetation; tends to saturate under dense canopy [85,86,87]

Comparative summary of vegetation indices applied in precision viticulture. Each index is described by its formula, spectral bands and ranges used, main viticultural applications, and specific limitations observed in vineyards. The table highlights the complementary role of NDVI, GNDVI, SAVI, NDRE, CWSI, and LAI in monitoring vine vigor, canopy structure, chlorophyll content, and water stress.

). NDVI remains the most widely used index due to its simplicity and broad applicability, whereas NDRE and GNDVI improve sensitivity to chlorophyll and nutrient dynamics under specific canopy conditions. SAVI is particularly valuable in early growth stages or vineyards with sparse canopy cover, while CWSI adds a thermal-based dimension for assessing water stress. The integration of LAI estimation further complements these indices by directly linking canopy architecture with grape yield and quality potential.

4.1. NDVI

The Normalized Difference Vegetation Index (NDVI), whose mathematical formulation presented in Table 1 is one of the most widely used indicators for evaluating vine vigor and vegetation vitality. This normalization compensates for variations in illumination, sensor sensitivity, and atmospheric conditions, allowing consistent comparison of vegetation vigor across different dates, flights, and sensors. NDVI exploits the contrasting spectral behavior of healthy vegetation, which absorbs light in the red band for photosynthesis and reflects strongly in the near-infrared (NIR) region due to leaf cellular structure [88]. As a result, high NDVI values indicate dense and physiologically active canopies, while low values are associated with stressed or sparse vegetation [89,90]. Typical NDVI vigor classifications are shown in

Table 2. NDVI classification ranges for vine vigor, derived from UAV multispectral imagery.

NDVI Values	Plant Status
−1–0	Dead Plant or Inanimate Object
0–0.33	Unhealthy Plant
0.33–0.66	Moderately Healthy Plant
0.66–1	Very healthy Plant

Source: Earth Observing System.

, based on UAV-derived multispectral imagery.

From a biophysical perspective, NDVI is valued for its simplicity and strong correlation with key biophysical traits, such as the leaf area index (LAI), yield potential, and grape quality parameters including sugar content and phenolic maturity [82]. For example, Matese et al. [91] reported that UAV-based NDVI reached a maximum R² = 0.61 (p < 0.001) in predicting total soluble solids (TSS) during véraison, highlighting its effectiveness in monitoring grape maturity at critical phenological stages. Despite these advantages, NDVI also presents notable limitations. Its values are highly influenced by environmental factors such as illumination, atmospheric variability, soil background, and shadowing [79]. Moreover, NDVI tends to saturate under dense canopies, reducing its sensitivity in high-vigor vineyards where subtle differences in vegetation may be masked [92].

In practical viticulture, however, NDVI remains a cornerstone of precision vineyard management. It serves as a robust variable for zoning and site-specific decision-making, supporting assessments of vine vigor, water status, and other ecophysiological traits. Filippetti et al. [80] found that NDVI-based vigor classes correlated significantly (p < 0.05) with grape and wine quality, reinforcing its value for vineyard zoning. Low-vigor zones—defined by lower NDVI values—typically produced grapes with higher sugar and anthocyanin content, resulting in wines with more intense fruity aromas and greater sensory quality. Similarly, Dorin et al. [93] demonstrated that NDVI zoning in cool-climate Riesling vineyards effectively differentiated wines with distinct sensory profiles, although they emphasized that NDVI performance is highly dependent on vineyard conditions and harvest year.

Consequently, while NDVI holds clear agronomic and enological value, its interpretation must be context-specific, accounting for phenological stage and canopy density. In practice, NDVI is often complemented by additional indices to overcome its limitations—such as NDRE (Normalized Difference Red-Edge Index) and EVI (Enhanced Vegetation Index) for dense canopies, GNDVI (Green NDVI) and VARI for early-season vigor detection, SAVI (Soil-Adjusted Vegetation Index) for soil-background correction, PCDI (Plant Cell Density Index) for canopy structure evaluation, and PRI (Photochemical Reflectance Index) for stress detection. Integrating these complementary indices enables more phenology-aware and canopy-sensitive analyses, improving the precision of vineyard zoning and the overall reliability of UAV-based viticultural monitoring.

4.2. GNDVI and SAVI

Among the most relevant NDVI-derived indices for vineyard monitoring are the Green Normalized Difference Vegetation Index (GNDVI) and the Soil-Adjusted Vegetation Index (SAVI). Both were developed to overcome specific limitations of the conventional NDVI, particularly those related to canopy density and soil background effects (see Table 1).

The GNDVI enhances sensitivity to chlorophyll concentration and foliar nitrogen content because chlorophyll absorbs less radiation in the green region than in the red [94]. This index typically ranges from −1 to 1, with positive values associated with vegetation and negative values with soil or water [95]. Its performance, however, may vary according to phenological stage and environmental conditions, showing reduced correlation before ripening and being affected by soil background or atmospheric scattering [81]. To mitigate these limitations, GNDVI is often used in combination with NDVI or other spectral indices, providing complementary information on canopy status and nutrient dynamics.

Empirical studies confirm its agronomic value. For instance, Ferro et al. [79] reported a strong correlation (R² = 0.839) with grape yield during ripening, which is highly significant (p ≤ 0.001), although its performance is lower at earlier phenological stages compared with NDVI.

SAVI, on the other hand, was specifically developed to minimize soil reflectance effects in areas with low canopy density [96]. It utilizes NIR and red bands, incorporating a soil adjustment factor (L), and is particularly useful in early growth stages or sparse canopies (Table 1). A typical value is L = 0.5 for moderate vegetation cover, while higher values are recommended in sparse canopies [97]. This correction improves vigor estimation accuracy in vineyards with exposed soil or inter-row vegetation, where soil background strongly affects reflectance. Consequently, SAVI enhances correlations with biophysical parameters such as leaf area, crop coefficient (Kcb), and yield, particularly under heterogeneous or arid conditions [98,99].

Although SAVI has been widely applied in crops such as corn, soybean, wheat, rice [100,101], and onion [102], its relevance in viticulture lies mainly in early growth stages, when canopy closure is incomplete. Campos et al. [103] demonstrated that integrating SAVI-derived Kcb values into evapotranspiration (ET) models slightly improved estimation accuracy, reducing RMSE by ≈2–5% (equivalent to ~0.1–0.3 mm day⁻¹) compared with constant Kc or NDVI-derived Kcb. Similarly, Qiao et al. [104]; highlighted SAVI’s ability to account for intra-canopy shading and variable soil exposure key factors in vineyards characterized by high structural heterogeneity.

In summary, both indices extend the applicability of NDVI to a broader range of vineyard conditions: GNDVI increases sensitivity to chlorophyll and nitrogen content, while SAVI minimizes soil background interference and improves precision under sparse or uneven canopies. Their usefulness remains context-dependent, and they are best regarded as complementary to NDVI and NDRE rather than direct replacements.

4.3. NDRE

The Normalized Difference Red Edge Index (NDRE) is a spectral index analogous to NDVI that relies on the near-infrared (NIR) band and a narrow spectral region between the red and NIR wavelengths, known as the “red edge” (Table 1). This band, typically spanning 712–722 nm, corresponds to the sharp transition in reflectance where chlorophyll absorption declines and internal leaf scattering increases. As a result, NDRE is particularly sensitive to subtle variations in chlorophyll concentration and canopy structure [82,105].

Compared with NDVI, NDRE shows greater sensitivity during mid and late growing stages, when high chlorophyll accumulation and canopy closure reduce NDVI responsiveness [106]. This enhanced sensitivity makes NDRE especially valuable for detecting physiological stress, nutrient status, and maturation dynamics in vineyards. For instance, Jorge et al. [83] reported a strong correlation between NDRE and SAVI (R² = 0.96) when assessing irrigation and water stress in grapevines, highlighting its reliability under sparse canopy conditions. In contrast, weaker correlations observed in olive orchards (R² = 0.17) illustrate that NDRE performance is strongly context-dependent, influenced by soil texture, canopy density, and vegetation cover.

Additional evidence from other crops reinforces its potential as a complementary indicator. Marty et al. [107] demonstrated that NDRE effectively captured variations in crop health and disease severity in lowbush blueberry fields, responding differently than NDVI depending on crop stage and management practices. In their study, NDRE values were obtained from UAV-based multispectral imagery acquired using a DJI Matrice 210 (DJI, Shenzhen, China) equipped with a MicaSense RedEdge-M camera (MicaSense Inc., Seattle, WA, USA). The UAV operated at an altitude of 120 m, achieving a ground sampling distance (GSD) of 8.2 cm px⁻¹, and captured imagery at one frame per second with 90% front and side overlap. The sensor covered blue, green, red, red-edge, and NIR bands with a horizontal field of view of 47.2°, parameters that ensured high spatial and radiometric quality essential for accurate NDRE estimation and reproducibility in precision agriculture. These findings suggest that NDRE should not be considered a replacement for NDVI but rather a complementary index providing additional insight into plant vigor and stress across phenological phases, particularly when integrated with complementary spectral and physiological indicators.

4.4. CWSI

The Crop Water Stress Index (CWSI) is a thermal-based indicator designed to quantify plant water stress by analyzing canopy temperature differences relative to well-watered and fully stressed reference baselines (Table 1) [108]. It is particularly relevant in viticulture due to its direct sensitivity to stomatal closure and transpiration dynamics under water deficit.

Values range from 0 (no stress) to 1 (maximum stress), providing a straightforward, scalable tool for assessing vineyard water status. Several studies have validated its utility. Matese et al. [51] reported strong correlations between CWSI and physiological traits such as photosynthesis, stomatal conductance, and chlorophyll fluorescence under regulated deficit irrigation. Alchanatis et al. [6] improved its accuracy by combining UAV thermal and multispectral data, filtering sunlit canopy pixels to reduce background interference. Similarly, López-García et al. [109] and Romero et al. [84] showed that UAV thermal imagery enhanced the spatial detection of stress zones, especially at midday when stomatal regulation peaks.

Comparative assessments highlight that CWSI can outperform multispectral indices such as NDVI or GNDVI for detecting real-time water stress, as it directly reflects stomatal physiology rather than indirect spectral responses. Möller et al. (2007) [110] found high correlations between CWSI and stomatal conductance (r = 0.91) and moderate correlations with stem water potential (Ψstem). Furthermore, UAV-based CWSI mapping has been successfully integrated into precision irrigation models to optimize water allocation and reduce resource waste. Cetin et al. [111] generated distributed CWSI maps using the METRIC method, achieving strong correlations (r = 0.95) with leaf area index (LAI). Recent advances in machine learning approaches have also been applied for automated stress classification from CWSI data, further supporting operational decision-making [112].

Despite these advantages, CWSI implementation requires rigorous calibration of wet and dry reference surfaces and remains sensitive to canopy architecture, atmospheric conditions, and image acquisition timing. These challenges can be mitigated by integrating thermal with multispectral data and using radiometrically calibrated sensors [51,84,113]. Overall, the strong physiological foundation and proven UAV compatibility of CWSI position it as one of the most reliable indicators of vine water status. Its ability to provide direct, spatially explicit insights into plant water relations confirms its value as a central tool for advancing precision irrigation and sustainable vineyard management [114].

4.5. LAI (Leaf Area Index)

In viticulture, the Leaf Area Index (LAI) has emerged as a key structural parameter that complements conventional vegetation indices, particularly when derived from UAV multispectral imagery. LAI, defined as the ratio of total leaf area to ground surface area (m² of leaves per m² of soil), provides crucial insights into canopy vigor, architecture, and photosynthetic capacity, all of which directly influence both grape yield and quality (Table 1) [103]. By capturing aspects of canopy structure that traditional indices like NDVI may overlook, LAI enables more precise vineyard management.

When using UAVs, LAI is typically estimated as the ratio between segmented canopy leaf area and projected vineyard ground area, where canopy segmentation defines the numerator and the denominator corresponds to the projected ground surface. Converting pixels to surface units through the Ground Sampling Distance (GSD) yields a dimensionless value (m²/m²) that is directly comparable across different vineyards and flight missions. Vélez et al. [86] demonstrated the validity of this approach by reporting a strong correlation (R² = 0.76, p < 0.01) between shaded canopy area and LAI using shadow analysis combined with k-means and Random Forest algorithms.

Recent studies highlight the robustness of UAV-derived LAI estimates. Ilniyaz et al. [87] demonstrated high accuracy (R² = 0.899, p < 0.05) using RGB and multispectral imagery, illustrating the effectiveness of conventional optical sensors for vineyard monitoring. Stobbelaar et al. [115] further enhanced LAI prediction by integrating thermal and VNIR data within a PLSR model (R² = 0.79, p < 0.001), particularly useful in dense canopies where NDVI tends to saturate. PLSR (Partial Least Squares Regression) reduces dimensionality by transforming collinear predictors into latent components, while VNIR (Visible–Near Infrared, 400–1000 nm) expands spectral information captured by UAV sensors. Although hyperspectral sensors are not yet widely applied in practical viticulture, experimental studies have shown promise. For example, Mesas-Carrascosa et al. [116] developed PLSR models (R² = 0.78) selecting key spectral bands across phenological stages, and incorporating land surface emissivity (LSE) from vegetation cover substantially improved prediction compared with spectral data alone.

Despite these advances, certain challenges remain. LAI estimation is sensitive to illumination, inter-row vegetation, and vineyard structures [85]. Nevertheless, UAV-based LAI retrieval provides a non-destructive, scalable, and reliable method for continuous canopy monitoring. This information supports precision management of canopy structure, irrigation, and fertilization, ultimately contributing to optimized yields and enhanced wine quality.

5. Applications of UAV-Based Indices

5.1. Water Stress and Irrigation

Water availability is one of the most critical environmental factors influencing grapevine physiology and grape quality, as it directly regulates photosynthesis, berry development, and the biosynthesis of key compounds that shape wine characteristic [19]. In many viticultural regions, vines rely on rainfall and groundwater; however, supplementary irrigation is often essential to maintain optimal water status. Water deficit triggers physiological adjustments such as reduced stomatal conductance, lower CO₂ assimilation, and impaired accumulation of sugars, aromatic precursors, and phenolic compounds, ultimately affecting grape composition and wine sensory attributes [20,117].

Traditionally, direct measurements such as stem water potential (Ψstem), stomatal conductance (gs), leaf turgor pressure, berry weight, and chlorophyll fluorescence have been used to quantify water stress in vineyards [47]. Although these approaches are reliable, they are destructive, labor-intensive, and spatially limited, which restricts their applicability for precision irrigation. In contrast, remote sensing technologies, particularly UAVs equipped with multispectral and thermal cameras, have emerged as scalable, non-destructive alternatives for detecting canopy-level stress responses. These systems capture physiological signals, including variations in reflectance and surface temperature linked to transpiration dynamics, enabling spatial mapping of vineyard water status [24,50].

Multispectral sensors capture reflectance in visible, near-infrared (NIR), and red-edge regions, as well as in shortwave infrared (SWIR), all of which are correlated with leaf structure, chlorophyll content, and water balance [51]. Thermal imaging, however, is particularly sensitive to transpiration, as stomatal closure under water stress elevates canopy temperature. Consequently, combining multispectral and thermal imaging provides more accurate mapping of vineyard variability, improving stress detection. Among thermal-based metrics, the Crop Water Stress Index (CWSI) has become one of the most robust tools. CWSI is calculated from the difference between canopy temperature and reference surfaces representing fully transpiring and non-transpiring leaves [118]. Its physiological relevance has been validated in multiple studies; for instance [51], demonstrated strong correlations of CWSI with photosynthesis, stomatal conductance, and chlorophyll fluorescence under regulated deficit irrigation, and highlighted the advantage of integrating thermal and multispectral data for more precise canopy mapping.

Likewise, Alchanatis et al. [6] showed that UAV-based thermal and multispectral integration improved CWSI map accuracy by minimizing background noise, while Gago et al. [50] and Tanda et al. [24] emphasized that combining CWSI with structural and spectral metrics enhances the detection of stress heterogeneity, providing a reliable proxy for vineyard water status.

Recent studies further confirmed the operational feasibility of UAV-based thermal imaging. Araújo-Paredes et al. [119] demonstrated that CWSI derived from UAV thermography strongly correlated with stem water potential in Loureiro vineyards, supporting its use as a practical monitoring tool. More recently, Lee et al. [120] reported that integrating thermal and multispectral data improved spatial detection of stress patterns within vineyard blocks, reinforcing the advantage of multisource approaches. On the other hand, Burchard-Levine et al. [121] applied UAV thermal imagery within Two-Source Energy Balance (TSEB) models to estimate evapotranspiration (ET) and developed new indices such as the Crop Transpiration Stress Index (CTSI) and Crop Stomatal Stress Index (CSSI). These indices provided stronger physiological relevance than traditional CWSI, particularly when partitioning canopy fluxes, highlighting the growing versatility of thermal approaches in viticulture.

Spectral indices remain widely used proxies of vineyard water status. NDVI, for example, has been correlated with stem water potential, stomatal conductance, and transpiration, providing useful though indirect insights into vine water balance [84,122]. It has also been applied to estimate crop coefficients (Kc) and delineate irrigation zones [123,124]. UAV-derived NDVI time series have proven effective for tracking intra-vineyard variability and evaluating regulated deficit irrigation strategies [53,125]. However, NDVI tends to saturate under high biomass and does not directly reflect water content, limiting its standalone utility. Complementary indices address these shortcomings [124]: demonstrated that GNDVI correlates with crop coefficients and yield, and that canopy area was an even better predictor of Kc than NDVI or GNDVI, underscoring the importance of integrating structural canopy traits with spectral indices. Similarly, NDRE, by incorporating the red-edge band, provides improved sensitivity during mid- to late-phenological stages, offering complementary information to NDVI in monitoring chlorophyll dynamics and water stress [83,126].

Reynolds et al. [127] confirmed significant correlations between NDVI and physiological traits such as Ψleaf and gs, reinforcing its role as a proxy when validated with field measurements. Similarly, studies [128] observed that NDVI and GNDVI correlated with stomatal conductance (r = 0.56–0.65) and yield (r = 0.68–0.73), while canopy temperature showed inverse correlations, confirming the value of combining spectral and thermal data. Beyond multispectral and thermal imagery, UAV-based RGB indices have been evaluated as low-cost alternatives. López-García et al. [109] demonstrated that indices such as the Green Leaf Index (GLI) and the Visible Atmospherically Resistant Index (VARI) correlated with accumulated water stress (SΨ), although these indices only performed reliably under moderate to severe stress, limiting their use for precision irrigation scheduling.

Machine learning approaches have also been integrated to improve predictive performance. Romero et al. [84] applied artificial neural networks (ANN) and generalized linear models (GLM) to estimate the Water Stress Index (ISW), combining vegetation indices with meteorological variables such as air temperature and vapor pressure deficit (VPD), which significantly improved stress prediction accuracy and underscored the value of multisource data integration. The interaction of multiple stressors further complicates vineyard monitoring. Cogato et al. [20] showed that high temperature and water deficit jointly influenced stomatal conductance, transpiration, and photosynthesis, altering spectral signatures captured by vegetation indices.

Overall, UAV-based sensing provides a comprehensive framework for water stress detection in viticulture, integrating spectral, thermal, and structural data to capture vine physiological responses. While traditional indices such as NDVI, GNDVI, and NDRE continue to offer valuable insights, their limitations highlight the need for complementary metrics such as CWSI and newly derived thermal indices. The integration of multisource data validated against field-based physiological measurements emerges as the most promising approach for supporting precision irrigation strategies and mitigating the impacts of water stress on grapevine productivity and wine quality. A consolidated comparison of these approaches, including key findings and limitations in vineyards, is presented in

Table 3. Remote sensing approaches applied for detecting water stress in vineyards using UAV-based multispectral and thermal sensors.

Index/Approach	Sensor Type	Key Findings in Vineyards	Limitations	Reference(s)
CWSI	Thermal	Strong correlation with stomatal conductance (r = 0.91) and Ψstem; reliable proxy of transpiration and water status; effective for regulated deficit irrigation	Requires calibration of wet/dry references; sensitive to canopy architecture and acquisition timing	[6,51,118,119,120]
NDVI	Multispectral	Correlated with Ψleaf, gs, transpiration, and grape yield; useful in mapping intra-vineyard variability and deficit irrigation strategies	Saturates under high biomass; indirect proxy of water status	[39,84,122,127]
NDRE	Multispectral	High sensitivity to chlorophyll concentration and mid-season stress; complementary to NDVI in ripening stages	Limited performance in sparse canopies and heterogeneous soils	[107,126]
GNDVI	Multispectral	Moderate correlation with crop coefficient (Kc, R2 = 0.36); useful for assessing nutritional and hydric stress	Sensitive to soil background; less robust at early phenological stages	[124]
RGB indices (GLI, VARI)	RGB UAV	Detected moderate-to-severe accumulated water stress (SΨ); low-cost alternative for monitoring	Low sensitivity to early or mild stress; limited use for irrigation scheduling	[109]
Integrated ML models (ANN, GLM)	Multispectral + weather data	Improved estimation of water stress index (ISW) by combining spectral and meteorological variables	Require site-specific calibration and large training datasets	[20,84]
TSEB-derived indices (CTSI, CSSI)	Thermal + energy balance models	Capture stomatal and transpiration stress with improved physiological relevance; better than empirical CWSI	Computationally complex; require partitioning of canopy fluxes	[121]

UAV-based spectral and thermal approaches for vineyard water stress detection. CWSI = Crop Water Stress Index; Ψstem = stem water potential; Ψleaf = leaf water potential; gs = stomatal conductance; Kc = crop coefficient; SΨ = accumulated water stress; ANN = Artificial Neural Network; GLM = Generalized Linear Model; TSEB = Two-Source Energy Balance model; CTSI = Crop Transpiration Stress Index; CSSI = Crop Stomatal Stress Index.

.

5.2. UAV for Pests and Diseases

Grapevines are highly susceptible to a wide range of pests and diseases that can substantially compromise both yield and grape quality. Traditional monitoring approaches, such as visual inspections and manual sampling, remain widely used; however, they are labor-intensive, spatially limited, and often subjective, relying heavily on the inspector’s expertise [129,130]. In the context of precision viticulture, early and objective detection of biotic stressors is crucial, yet it remains challenging. Remote sensing technologies, particularly UAV-based multispectral and RGB imaging, have emerged as powerful alternatives, offering non-destructive, rapid, and spatially explicit assessments of plant health [131]. Pathogen infection alters chlorophyll absorption and canopy reflectance properties, enabling the use of vegetation indices to discriminate between healthy and diseased vines. Numerous studies have demonstrated the applicability of UAV-derived indices in detecting major grapevine diseases, although their effectiveness is often disease- and context-specific. For instance, NDVI and related indices have been widely applied to fungal diseases such as grapevine leaf stripe disease (GLSD) [44], powdery mildew [45], and Botrytis bunch rot [132]. While these indices allow detection of symptomatic vines, their capacity to identify early or mild infections is limited, as NDVI often saturates under dense canopies or fails to differentiate between abiotic and biotic stressors. Similar limitations have been reported for Esca, where NDVI successfully identified severely infected vines but showed poor sensitivity in early-stage infections [133].

Viral and vector-transmitted diseases, such as Grapevine Leafroll-Associated Virus (GLRaV) and Flavescence Dorée (FD), pose additional challenges. Reynolds et al. [127] demonstrated that NDRE and REIP were more effective than NDVI for GLRaV detection, reflecting changes in pigment composition and photosynthetic performance. In the case of FD, other studies [134,135] reported near-perfect classification accuracies (>94%) in red cultivars using indices such as GRVI and RGI, whereas accuracies in white cultivars were considerably lower due to less pronounced symptom expression. These findings highlight a critical limitation: varietal differences strongly influence the diagnostic performance of spectral indices.

Pest detection presents further complexity. UAV-based multispectral and hyperspectral imaging has been used to detect phylloxera infestations by identifying altered reflectance signatures and reduced vigor [136]. However, distinguishing phylloxera-induced stress from other stressors remains difficult, underscoring the risk of false positives. Similarly, studies [137] applied UAV-derived indices (ExGR and GNDVI) to detect Cynodon dactylon infestations, achieving > 97% accuracy and enabling optimized herbicide application. While promising, these approaches primarily address weed mapping rather than direct pathogen detection, indicating the need for stressor-specific strategies.

Collectively, these studies confirm the potential of UAV-based spectral imaging for vineyard pest and disease monitoring, while also revealing important limitations. Index-based approaches are highly dependent on canopy structure, environmental conditions, and the phenological stage of the crop, constraining their universal applicability. Moreover, spectral responses to biotic stress often overlap with those caused by abiotic stressors such as nutrient deficiencies or water stress, limiting specificity. Advanced methods including optimal band selection, multi-sensor integration (e.g., thermal + spectral), and machine learning algorithms are increasingly necessary to improve diagnostic accuracy and enable robust operational deployment. Nevertheless, UAV-based imaging remains a valuable complement to traditional scouting methods, enhancing the spatial and temporal resolution of vineyard monitoring and supporting more efficient and sustainable disease management strategies (

Table 4. UAV-based spectral indices and imaging approaches for the detection of pests and diseases in vineyards.

Disease/Pest	Sensor Type	Index/Approach	Key Findings in Vineyards	Limitations	Reference(s)
Grapevine Leaf Stripe Disease (GLSD)	Multispectral	NDVI	Differentiated symptomatic from healthy vines at canopy level	Limited in detecting early/asymptomatic infections	[44]
Powdery mildew (Uncinula necator)	Multispectral	NDVI	Strong correlation (r > 0.9) with disease severity under field conditions	Symptom expression varies across cultivars; canopy shading reduces accuracy	[45,138]
Grapevine Leafroll Virus (GLRaV)	Multispectral	NDVI, NDRE, REIP	NDRE and REIP improved virus detection compared to NDVI alone; linked to pigment changes	Variable accuracy between vineyards; canopy structure affects detection	[127]
Botrytis bunch rot	Multispectral + RGB	NDVI	Early signs detected through reflectance differences	Overlaps with abiotic stress; low specificity	[128]
Esca (Trunk disease)	Multispectral	NDVI, GNDVI	Diseased vines had consistently lower NDVI (0.68–0.79) vs. healthy vines	Ineffective at detecting mild/early infections	[133]
Phylloxera	Multispectral + Hyperspectral + RGB	Vegetation indices (MCARI, Red-edge indices)	Detected spectral traits linked to reduced vigor and chlorophyll	Confounded with abiotic stress; requires hyperspectral data	[136]
Flavescence Dorée (FD)	RGB + Multispectral	GRVI, RGI, NDVI, CI	High discrimination in red cultivars (AUC ≈ 1.0); band selection at 520–800 nm improved classification (>94%)	Accuracy lower in white cultivars; requires optimized sensor settings	[134,135]
Jacobiaska lybica (mealybug vector)	RGB	RGB indices	Effective mapping of symptomatic patches	Poor generalization to early stages; manual validation required	[18]
Cynodon dactylon (weed competition)	RGB + RGB-NIR	ExGR, GNDVI	Differentiated weeds from soil with >97% accuracy; enabled 48% reduction in herbicide use	Not specific to disease; sensitive to soil background	[137]

UAV-based multispectral and RGB indices applied to the detection of major pests and diseases in vineyards. NDVI = Normalized Difference Vegetation Index; NDRE = Normalized Difference Red Edge; REIP = Red Edge Inflection Point; GRVI = Green-Red Vegetation Index; RGI = Red-Green Index; CI = Chlorophyll Index; MCARI = Modified Chlorophyll Absorption in Reflectance Index; ExGR = Excess Green-Red Index.

).

5.3. UAV Applications in Grape Ripening and Plant Physiology

Grape ripening represents one of the most decisive phases in the winemaking chain, as it determines the biochemical composition of berries and, consequently, the sensory and aging potential of wines. During ripening, sugars progressively accumulate, acidity decreases, and secondary metabolites such as phenolics, anthocyanins, and aromatic precursors are synthesized, all of which directly shape wine quality [139,140]. From an enological standpoint, harvesting at the optimal phenological stage is essential, since misaligned timing may result in wines with excessive alcohol, poor acidity, or weak aromatic expression [49,141]. The concept of “industrial maturity,” defined by the stabilization of the sugar-to-acidity ratio, remains central to balancing wine alcohol content, pH, and organoleptic quality [48].

However, the main challenge for viticulturists is not only to determine the average maturity of a vineyard, but also to capture its spatial heterogeneity, which is often exacerbated by soil variability, microclimatic gradients, or rootstock cultivar interactions [13,60]. Ripening is orchestrated by a tightly regulated sequence of physiological processes that alter both berry composition and plant optical properties. Vine water status, phloem unloading dynamics, and potassium accumulation strongly influence sugar partitioning and berry pH regulation [80,142]. The véraison stage marks a physiological turning point, characterized by berry softening, color change, and phenolic reprogramming, accompanied by rapid changes in berry optical reflectance [143]. These transitions provide the scientific basis for remote sensing, since UAV-based multispectral imagery can detect subtle physiological changes that drive enological potential [144].

Traditional maturity monitoring relies on destructive sampling of berries to measure soluble solids, titratable acidity, and pH. While analytically precise, these methods are labor-intensive, spatially restricted, and poorly suited to capture heterogeneity across vineyard blocks [16]. This limitation is magnified under climate change, where asynchronous ripening and uneven sugar-to-acidity ratios challenge uniform harvest planning [140]. UAV-based multispectral sensing overcomes these constraints by offering non-destructive, repeatable, and spatially explicit monitoring of ripening gradients, allowing the identification of “hot spots” of early or delayed maturity [53,93]. Such spatial insights facilitate selective harvesting and zonal winemaking, bridging plant physiology with enological outcomes.

Vegetation indices (VIs) derived from UAV imagery are among the most widely applied tools for assessing ripening. NDVI remains the benchmark index, with numerous studies showing its correlation with °Brix, acidity, anthocyanins, and color [20,145]. However, NDVI saturation under high biomass limits its precision during advanced stages of ripening [13]. Alternative indices provide improved sensitivity: GNDVI is more responsive to chlorophyll and nitrogen status, enhancing yield and crop coefficient estimation [124]; NDRE is particularly effective during mid-to-late ripening, capturing subtle chlorophyll declines linked to maturity [126]; while PRI has been associated with photosynthetic efficiency and stress responses, enabling the interpretation of ripening under environmental constraints [105,146]

Composite indices such as TCARI/OSAVI have further refined ripening assessment by enhancing sensitivity to carotenoid and chlorophyll variations, particularly under mild stress, thereby linking ripening to water balance [146]. Red-edge indices have also been linked to phenolic and flavonoid accumulation, reinforcing their role as biochemical proxies [147,148,149]. Ferrer et al. [150] demonstrated significant associations between NDVI-derived vigor classes and grape juice acidity, phenolic index, and leaf area, while Dorin et al. [93] showed that vigor zoning predicted distinct enological outcomes in Riesling zonal vinification. These studies highlight that UAV spectral indices not only track physiological ripening, but can also predict final wine composition.

More recently, RGB-based indices such as RGBVI2 and RGBVI3, derived from standard UAV RGB cameras, have emerged as cost-effective alternatives for small-scale growers [68]. Although less physiologically precise than multispectral sensors, they democratize vineyard monitoring, particularly when multispectral or hyperspectral systems are economically prohibitive.

Beyond phenolics and sugars, UAV indices have been associated with macro- and micronutrient dynamics. Peng et al. [151] predicted leaf nitrogen, phosphorus, and potassium from UAV multispectral data, linking nutrient availability to berry composition and fermentation potential. Taskos et al. [147] confirmed correlations between red-edge indices and phenolic content, highlighting nutrition as a hidden driver of wine variability. Together, these findings underscore the multi-dimensional role of UAV sensing in monitoring both ripening physiology and nutritional determinants of wine quality.

Despite these advances, significant knowledge gaps remain. First, spectral indices capture canopy traits indirectly related to berry metabolites, and correlations often vary by cultivar, terroir, and season [13,140]. Standardized calibration frameworks integrating UAV indices with biochemical assays are needed to improve transferability. Second, most studies provide single-time measurements at véraison or harvest, neglecting the dynamic trajectory of ripening. High-frequency UAV monitoring across phenological stages would enable better predictions of optimal harvest windows and enhance selective harvesting strategies. Third, NDVI and related indices suffer from saturation and environmental dependencies (light conditions, canopy geometry, soil background), limiting universal applicability. Integrating spectral data with thermal imagery (water status) and 3D canopy traits such as leaf area or pruning weight [68] offers a more holistic approach.

Emerging advances in machine learning further expand UAV applications in viticulture. Recent studies have demonstrated that ML models integrating UAV spectral data can significantly improve predictions of grape yield and phenolic composition [116]. Similarly, other works have shown that combining UAV-derived indices with phenological information enhances vineyard performance forecasting [152]. In addition to empirical vegetation indices, mechanistic and hybrid modeling approaches have been increasingly explored to estimate grapevine physiological parameters and ripening dynamics [153]. Mechanistic models simulate biophysical and biochemical processes such as photosynthesis, transpiration, and sugar accumulation, thereby linking canopy structure and water status with berry composition [79,154].

A primary application of UAV-based plant phenotyping is to provide the high-resolution, spatially explicit data required to calibrate and validate these mechanistic models, moving them from theoretical frameworks to vineyard-specific tools. Building on this, hybrid models integrate UAV-derived spectral, thermal, and structural phenotyping data with mechanistic frameworks or machine learning algorithms, improving predictive accuracy and enabling spatially explicit assessments of vineyard heterogeneity [155]. These approaches facilitate dynamic monitoring across phenological stages, support the prediction of optimal harvest windows, and provide a more robust interpretation of ripening progression by combining process-based understanding with high-resolution remote sensing observations [95].

Future research is expected to further expand these hybrid frameworks by incorporating climatic, soil, and management variables, thereby consolidating UAV-based phenotyping as a powerful decision-support tool for precision viticulture [156]. Collectively, these results suggest that the future of UAV sensing lies in the development of predictive, decision-support systems that integrate multisource datasets, including spectral, thermal, structural, climatic, and agronomic variables. As summarized in

Table 5. Studies relating UAV-based vegetation indices with grape ripening and wine quality parameters.

Index	Sensor	Key Findings in Vineyards	Limitations	Reference
NDVI, LAI	Multispectral	NDVI explained phenolic variability across vineyard blocks	Requires calibration with destructive sampling	[82]
NDVI	Multispectral	Strong correlation between NDVI and berry ripening parameters	NDVI saturates at high vigor	[150]
NDVI, NDRE	Multispectral	Red-edge indices improved sensitivity to chlorophyll/nutrient status	Sensitive to soil/background effects	[147]
NDVI	Multispectral, multiple	NDVI captured anthocyanin accumulation trends	Limited resolution at late ripening stages	[148]
NDVI	Multispectral	UAV-NDVI predicted °Brix variability	Dependent on local calibration	[157]
NDVI	Multispectral	Seasonal correlation between NDVI and berry composition	Correlations varied across seasons	[13]
NDVI	Multispectral	Early evidence linking canopy vigor to wine color	Did not include UAV imagery	[145]
TCARI/OSAVI	Multispectral + RGB + NIR	Combined indices improved detection of chlorophyll/carotenoids	Sensitive to light conditions	[146]
NDVI-RGBVI	Multispectral-RGB	Low-cost RGB indices captured ripening trends	Overlap with vigor effects Effective only at moderate–high stress	[149]
NDVI, OSAVI, MSAVI, MCARI	Multispectral	Nutrient status strongly linked with ripening (N, P, K)	Requires simultaneous leaf sampling	[151]
VARI, PRI, RGBVI	RGB	Detected phenolic maturity and malic acid changes	RGB less reliable under canopy shadow	[68]
NDVI, GNDVI, NDRE, MSAVI	Multispectral	Strong correlations with berry composition and yield	Cultivar-specific responses not standardized	[79]

UAV-based vegetation indices applied to grape ripening and wine quality assessment. Key findings highlight correlations between NDVI, red-edge, and composite indices with berry composition, nutrient status, and phenolic maturity. Limitations include NDVI saturation, soil background sensitivity, dependence on calibration, and cultivar-specific variability.

, UAV-based vegetation indices have proven effective for monitoring grapevine physiology, ripening progression, and yield estimation. However, the path forward requires standardized calibration protocols, multi-temporal monitoring, multisource integration, and AI-based modeling to fully bridge vineyard heterogeneity with production outcomes. Progress in these areas will consolidate UAVs as essential tools for precision viticulture, directly linking ripening physiology with vineyard yield.

6. Limitations

The use of UAVs equipped with multispectral cameras has significantly advanced vineyard monitoring, enabling detailed assessments of canopy vigor, water status, and spatial variability. Nevertheless, several technical and operational limitations still affect the accuracy, reproducibility, and comparability of results across studies.

First, one of the most persistent challenges lies in achieving robust radiometric calibration. Fluctuations in solar irradiance, sensor sensitivity, and atmospheric conditions can substantially alter reflectance values and, consequently, the vegetation indices derived from them. Correction protocols such as reflectance panels, incident light sensors (ILS), and the Empirical Line Method (ELM) are commonly applied to mitigate these effects. However, the limited transparency of the algorithms integrated into commercial processing platforms often restricts methodological reproducibility and complicates cross-study comparisons under different vineyard conditions or sensor configurations.

In addition, geometric distortions pose a considerable obstacle, particularly when UAV flights rely on inaccurate GPS positioning or a suboptimal distribution of ground control points (GCPs). Typically, a minimum of 5–10 well-distributed GCPs per hectare is recommended to ensure accurate georeferencing and orthomosaic generation. In such circumstances, spatial errors propagate through orthomosaic, compromising canopy segmentation and vegetation index accuracy. To overcome this issue, the adoption of Real-Time Kinematic (RTK) or Post-Processed Kinematic (PPK) GNSS systems, combined with robust photogrammetric workflows and optimized flight planning strategies, has proven effective in enhancing spatial fidelity and reducing positional uncertainty [158,159].

Moreover, environmental conditions during image acquisition strongly influence data quality. Wind may compromise flight stability, reduce image sharpness, and even disrupt data transmission, especially in consumer-grade UAVs [160,161,162]. Similarly, variability in illumination driven by solar angle, cloud cover, or terrain-induced shadows reduces spectral consistency between flights. To mitigate these factors, corrections based on Digital Elevation Models (DEM) have been recommended [32,152].

Equally, the timing of image acquisition plays a decisive role in ensuring reliable monitoring. Several studies indicate that midday, particularly between 12:00 and 13:30, is the most favorable interval to capture thermal and multispectral images, since the influence of shaded leaves and stomatal closure is minimized [6,163]. Furthermore, even a one-hour difference can introduce substantial variability in vegetation indices, which highlights the need for consistent scheduling of UAV flights to guarantee temporal comparability [84].

Finally, spectral contamination caused by inter-row vegetation, bare soil, or canopy shadows often complicates the interpretation of vegetation indices. Although advanced segmentation and classification techniques such as CLARA, k-means, Random Forest, and deep learning methods have shown promising results in isolating vine canopy from surrounding elements, their performance remains sensitive to vineyard heterogeneity and complex canopy architecture [39,72,164].

7. Future Projections for UAV Implementation

The adoption of UAVs in agriculture is projected to expand steadily in the coming years, largely because of their ability to reduce operational costs, enhance productivity, and support more sustainable practices. Nevertheless, investment in high-quality platforms equipped with multispectral sensors remains substantial, and recurrent expenses for calibration, maintenance, and upgrades continue to limit access for many growers. Skilled personnel are also required to ensure accurate flight planning, image acquisition, and data interpretation, which poses challenges for small and medium-scale vineyards.

A promising avenue for future research is the integration of UAV-derived spectral information with complementary technologies, including soil moisture sensors, automated irrigation systems, and climate-responsive decision-support tools. Such integration could connect vigor maps with real-time environmental data, refining harvest scheduling, canopy regulation, and irrigation efficiency. Beyond viticulture, drones are increasingly contributing to environmental monitoring of water resources, soil health, and ecosystem services, consolidating their role as versatile instruments across agricultural systems.

Recent progress in artificial intelligence suggests that UAVs may evolve from monitoring tools into predictive systems. By combining spectral indices with climatic variables, 3D canopy structure, and biochemical assays, it may become possible to estimate phenolic composition, acidity, or technological maturity with a level of accuracy relevant to enological decision-making. Yet a major gap remains: while correlations between NDVI, NDRE, or GNDVI and berry physiology are well documented, the standardized connection with final wine attributes protein stability, turbidity, phenolic profile, or sensory expression remains poorly established. Addressing this gap will require cross-regional calibration protocols and longitudinal studies capable of linking vineyard heterogeneity to wine quality outcomes.

Another critical step is to overcome existing technical constraints. Radiometric calibration procedures need to be harmonized; cost-effective RGB indices must be further validated; and advances in cloud-based platforms should facilitate data sharing and reproducibility across different terroirs. At the same time, regulatory frameworks governing UAV operations must adapt to accommodate wider adoption, balancing aerial safety with the management of increasingly complex datasets.

8. Conclusions

The implementation of unmanned aerial vehicles (UAVs) equipped with multispectral and thermal sensors has fundamentally transformed precision viticulture, providing a powerful, non-destructive, and spatially explicit approach for monitoring grapevine physiological status. This critical review has integrated and evaluated the current state of knowledge on UAV-derived vegetation indices, with particular emphasis on NDVI, NDRE, GNDVI, SAVI, and the thermal index CWSI, highlighting their operational strengths, limitations, and potential applications across diverse viticultural contexts.

NDVI remains the most widely used vegetation index due to its computational simplicity, robustness, and strong association with canopy vigor and productivity. However, its tendency to saturate in dense canopies and its sensitivity to soil background and shadows underscore the need to complement it with other indices for advanced applications. NDRE and GNDVI, by incorporating red-edge and green bands, respectively, exhibit superior sensitivity for estimating chlorophyll content and nutritional status, particularly during advanced phenological stages such as véraison and ripening. SAVI, through its soil-adjustment factor, proves especially valuable in young vineyards or sparse canopies where bare soil influences spectral reflectance. Meanwhile, CWSI, derived from thermal imagery, emerges as an indispensable tool for directly assessing water stress through physiologically relevant parameters such as stomatal regulation and transpiration dynamics.

The future of precision viticulture lies in the synergistic integration of spectral, thermal, and structural information with advanced data analytics. Combining multispectral and thermal indices with three-dimensional canopy models and machine learning algorithms will not only enhance monitoring but also enable predictive modeling of key grape and wine quality parameters, including sugar content, acidity, anthocyanins, and phenolic composition. Nevertheless, several technical and methodological challenges remain. Consistent radiometric calibration, correction for geometric and environmental effects, and the development of standardized acquisition and processing protocols are essential to ensure reproducibility and comparability across vineyards, cultivars, regions, and growing seasons.

By addressing these limitations and adopting rigorous validation strategies, UAV-based vegetation indices can evolve from experimental tools to operational decision-support systems. Their integration into vineyard management will foster more precise, sustainable, and quality-oriented viticulture, bridging the gap between vineyard variability and enological outcomes. In doing so, UAV technologies and vegetation indices will serve as foundational elements for the continued advancement of precision agriculture within the wine industry.