The Application of Amino Acids as a Sustainable Strategy for Managing Water Stress in Vineyards

Highlights

What are the main findings?

• UAV-derived multispectral and thermal indices successfully captured grapevine vigor and water stress responses to amino acid-based biostimulants under contrasting drought conditions.
• Foliar application of pyroglutamic acid was associated with higher yield components and more stable canopy thermal response in non-irrigated vines, particularly under moderate drought.

What are the implications of the main findings?

• Pyroglutamic acid shows potential as a complementary drought-mitigation strategy alongside irrigation management in water-limited viticulture systems.
• The NGRDI index, derived from low-cost RGB imagery, emerged as a promising indicator for vineyard water status monitoring in precision viticulture.

Abstract

Water scarcity increasingly threatens viticulture in the Macaronesian region due to climatic variability and recurrent droughts. This study evaluated the physiological and productive responses of grapevines (Vitis vinifera L.) to foliar applications of two amino acid-based biostimulants, pyroglutamic acid and pipecolic acid, under contrasting water availability conditions on Madeira Island, Portugal. Three non-irrigated treatments were arranged in a randomized complete block design: T1 (no irrigation and no amino acids), T2 (pyroglutamic acid, without irrigation), and T3 (pipecolic acid, without irrigation), while conventional irrigation (T4) was included as a non-randomized reference. Agronomic parameters and UAV-derived multispectral and thermal data were analyzed during the 2023 (moderate drought) and 2024 (severe drought) growing seasons. Vegetation indices (NDVI, GNDVI, NDRE, NGRDI, and GLI) and the Simplified Crop Water Stress Index (CWSI_si) were used to assess canopy vigor and plant water status. In 2023, T4 showed significantly higher bunch number and total yield, whereas differences among non-irrigated treatments were not statistically significant. Nevertheless, T2 showed consistent numerical trends toward higher yield components and a comparatively more stable canopy thermal response than the untreated control. In 2024, severe drought reduced productivity across all treatments, with no significant difference detected. Yield components were generally strongly correlated, while CWSI_si was negatively associated with vegetation indices, particularly under moderate drought. The NGRDI demonstrated potential as a low-cost RGB-based indicator but requires cautious interpretation. Overall, pyroglutamic acid may represent a complementary strategy to irrigation and UAV-based precision monitoring in drought-prone viticulture, although confirmation through longer-term and higher-powered field studies is required.

1. Introduction

In 2023, approximately 7.2 million hectares of land worldwide were under vineyards, according to the Organisation Internationale de la Vigne et du Vin—OIV [1]. The global wine production is 237 million hectoliters. France, Italy, and Spain together account for half of the world’s wine production [1]. Portugal’s wine production in 2024 is expected to reach 6.9 million hectoliters, which is 8% less than that in 2023. However, this amount is close to Portugal’s five-year average production, representing a small increase of 0.2%. Climate change makes grape growth more difficult because of rising temperatures, which change rain patterns and increase the frequency of extreme weather events such as droughts and heatwaves [2,3]. Although vines can handle dry conditions, they still require a large amount of water during grape growth, particularly during the driest months [4].

The increase in temperature is accompanied by a decrease in moisture in the atmosphere, which increases water loss from the soil and increases plant water demand to ensure the processes of evaporation and transpiration, respectively. According to Iglesias & Garrote [5], climate change will likely lead to spatially heterogeneous changes in precipitation across Europe, with increases in the north and decreases in the southern and Mediterranean regions, which may exacerbate water scarcity issues in agriculture. As a result, viticulture requires more water, making irrigation important to keep the vine healthy and prevent stress, especially in the southern wine regions [6]. Previously, European vineyards relied on rain, and high-quality grapes were often grown without irrigation [7]. However, with less water available in Southern Europe and other wine areas, the need for irrigation in vineyards has increased. This change has been significant over the past 20 years [7].

Since the 15th century, grape production has been important to the Madeira economy, culture, and identity. The island’s weather and soil conditions, with median temperatures, different microclimates, altitudinal gradients, and terraced vineyards, are important factors in grape growing and production [8,9]. Vineyards at higher altitudes, such as those in São Vicente and Calheta, receive different amounts of sunlight and temperatures, resulting in fresh, acidic, and complex wines. Madeira is famous worldwide for its fortified wine, owing to its unique conditions. The island’s nutrient-rich volcanic soils are ideal for vine growth and high-quality grape production. However, grape growth in Madeira is a significant challenge. Climate change, with more extreme events such as droughts, threatens both the quantity and quality of grapes [8]. In addition, the lack of flat land due to the island’s mountains limits the size and location of grape-growing areas.

Drought stress significantly reduces global crop production. This negatively affects the growth and yield of many important crops [10,11]. Drought, owing to water scarcity, limits nutrient uptake and gas exchange, which inhibits photosynthesis, slows down shoot growth, and reduces yield. It also causes metabolic and physiological changes in plants, affecting key processes such as respiration, enzyme activity, and chloroplast function [12,13]. Grapevines (Vitis vinifera L.) are crops that can grow and adapt to various climates. However, despite this adaptability, drought can adversely affect crop development and yield.

Therefore, finding ways to reduce drought stress is important, and a range of organic products, known as biostimulants, are now available on the market to promote sustainable agricultural practices. According to the European Biostimulant Industry Council, plant biostimulants contain substances and/or microorganisms that activate natural processes when applied to plants or their rhizospheres [14]. These processes enhance nutrient uptake and efficiency, improve crop quality, and increase abiotic stress tolerance [15].

Today, many products called biostimulants are available to farmers for use. These products help make farming more sustainable. They contain substances or microorganisms that, when used on plants or soil, help plants absorb nutrients, grow better, and cope with stress [15]. Recent studies have shown that they protect plants from drought, salinity, and extreme temperature [16,17,18]. Biostimulants help plants grow and produce more by improving stress responses and increasing antioxidant levels. They also help improve crop yield and quality by making nutrients more available and assisting soil processes [19].

Among the various biostimulant treatments, amino acids are frequently employed to enhance tolerance to water deficit. Notably, pyroglutamic acid has been shown to effectively induce drought tolerance in a range of crops under field conditions [20,21]. Similarly, pipecolic acid has been reported to enhance drought tolerance by strengthening antioxidant and stress-related responses [22].

Amino acids play a central role in plant responses to abiotic stress by acting as metabolic intermediates, osmoprotectants, antioxidants, and signaling molecules. Under water deficit, several amino acids contribute to osmotic adjustment, stabilization of cellular membranes, regulation of redox homeostasis, and modulation of stress-responsive pathways, ultimately supporting photosynthesis and growth maintenance [21,23]. In grapevine, amino acid metabolism has been associated with stress acclimation processes, including improved antioxidant capacity, regulation of stomatal behavior, and protection of the photosynthetic apparatus under drought conditions [11,24].

Among specific amino acids, pyroglutamic acid has received increasing attention as a biostimulant due to its role in the γ-glutamyl cycle and its involvement in nitrogen metabolism, antioxidant regulation, and osmotic balance. Field and controlled-environment studies have demonstrated that foliar application of pyroglutamic acid enhances drought tolerance by maintaining photosynthetic activity, improving water status, and reducing oxidative damage in crops such as lettuce and tomato [20,21]. Conversely, pipecolic acid is recognized as a key signaling molecule involved in stress-induced metabolic reprogramming and systemic acquired resistance, with documented roles in enhancing antioxidant responses and stress tolerance under water deficit in herbaceous crops [22,25]. Despite these advances, comparative field-based evidence on the physiological and productive effects of these amino acids in grapevine remains limited, particularly under contrasting irrigation regimes. This knowledge gap underpins the rationale for evaluating their potential as biostimulants for drought mitigation in viticultural systems.

However, biostimulants are most valuable when integrated with water-saving management, particularly in regions facing recurrent drought. In viticulture, this integration is increasingly achieved through precision irrigation strategies that match water supply to crop demand and phenology. UAV-based remote sensing, especially thermal imaging combined with spectral vegetation indices, provides a rapid and spatially explicit way to monitor canopy vigor and plant water status and to support irrigation decisions [26,27]. The Crop Water Stress Index (CWSI) proposed by Idso et al. [28] and subsequently refined by Jackson and authors [29] has become a widely used framework to interpret canopy temperature as an indicator of vine water stress, enabling more timely and efficient irrigation scheduling under water-limited conditions. Jackson and authors [29] improved this methodology by incorporating detailed meteorological data to strengthen the interpretation of canopy temperature-based stress signals.

This study evaluated the physiological and productive responses of grapevines to foliar applications of pyroglutamic acid and pipecolic acid under contrasting irrigation regimes, integrating UAV-based thermal imagery (CWSI_si) and vegetation indices to monitor canopy vigor and water stress. Because grapevine growth and canopy structure can vary substantially between seasons even under similar management, this two-season field experiment should be interpreted as an assessment of the potential of this approach rather than a definitive long-term conclusion. Longer multi-year monitoring is warranted to better separate interannual vine variability from treatment effects.

We hypothesized that foliar application of amino acid-based biostimulants (pyroglutamic acid and pipecolic acid) could modulate grapevine physiological responses to water deficit, leading to detectable differences in canopy water status and vigor indicators derived from UAV imagery. We further expected that agronomic responses would be more difficult to detect statistically in a field setting due to high inter-vine variability and the short duration of the trial.

2. Materials and Methods

2.1. Study Area and Experimental Design

The field trial was conducted at Quinta das Vinhas, Estreito da Calheta, Madeira Island, Portugal (Figure 1), in the vineyard of the Verdelho cultivar. The experimental site covered an area of 2.37 ha, with an altitude ranging from 305 to 347 m above sea level [8]. According to the Köppen–Geiger climate classification, the region is characterized as Csb, denoting a temperate climate with hot and dry summers.

The experiment followed a randomized complete block design (RCBD) with three treatments as follows: T1 (without irrigation and amino acids), T2 (Pyroglutamic acid, without irrigation), and T3 (Pipecolic acid, without irrigation), each replicated three times. Five vines were evaluated for each replicate. The replicates were distributed across three experimental rows of the field. A fourth treatment (T4), corresponding to conventional irrigation management, was included for comparative purposes. This treatment was not part of the RCBD and was applied without randomization in a single cultivation row with three replications, with five vines evaluated per replicate.

The conventional irrigation treatment (T4) followed the grower’s standard empirical irrigation practice, without sensor-based scheduling, regulated deficit irrigation, or real-time adjustment based on plant or soil water status. Irrigation management varied according to the treatments. In the three experimental rows (RCBD), irrigation was fully suspended before flowering, specifically on 2 May 2023 (1st crop cycle) and 27 May 2024 (2nd crop cycle). The fourth row followed a conventional irrigation schedule throughout the crop cycle and served as a fully irrigated control. This design enabled us to evaluate biostimulant treatments under both irrigated and non-irrigated conditions, simulating water stress scenarios aligned with projected climate scenarios and their impacts on island viticultural systems.

Irrigation management differed between treatments. In the experimental plots (T1–T3), irrigation was completely suspended after the first application of biostimulants in both growing seasons, to impose water deficit conditions and to assess plant responses under non-irrigated scenarios. In contrast, the irrigated treatment (T4) followed the farmer’s conventional irrigation practices and was maintained according to the grower’s empirical criteria, reflecting standard vineyard management in the region. No quantitative measurements of irrigation volume, crop evapotranspiration, or soil moisture were performed during the study, and irrigation in T4 was not scheduled based on sensor-driven or model-based approaches. Consequently, T4 was used as a qualitative reference for irrigated conditions rather than as a precisely controlled irrigation treatment.

2.2. Preparation and Application of Amino Acids

Pyroglutamic and pipecolic acid solutions were freshly prepared before each application by dissolving the compounds in distilled water to obtain a final concentration of 5 mM. Five liters of solution were prepared to ensure homogeneity and consistency of application across all replicates of each treatment. The initial application of amino acids was carried out on 2 May 2023 and 25 May 2024 (Full Flowering stage—BBCH 65). Subsequent reinforcement treatments were performed on 19 June 2023 and 25 June 2024 (Berries Pea-size—BBCH 75).

Foliar applications were performed at the phenological stages indicated above, according to the BBCH scale. Amino acid solutions were applied using a manual backpack sprayer, uniformly targeting the entire vine canopy. Spraying was conducted until full leaf wetting was achieved, ensuring homogeneous coverage of the foliage. No fixed spray volume per vine or per row was quantified, as applications were standardized based on visual assessment of complete leaf coverage rather than on volumetric dosage. Although a fixed spray volume per vine was not quantified, applications were standardized by applying the solution until full leaf wetting using the same operator, equipment, concentration, and application protocol across all treatments and replicates. This approach minimized operational variability and ensured comparable exposure among vines.

2.3. Climate Data

Local meteorological data were collected between January and August of each year to characterize the climatic conditions during the grapevine growth and development cycles in 2023 and 2024 (32.73°N, 17.19°W). The dataset included the average, maximum, and minimum air temperatures, as well as total monthly precipitation. These parameters are presented in Figure 2 and Figure 3, providing essential context for interpreting crop physiological and agronomic responses under varying environmental conditions. Cumulative precipitation patterns for each growing season are described in the text to contextualize drought timing, even though only monthly totals are shown in Figure 2 and Figure 3.

The meteorological data were obtained from a weather station installed within the study area, located approximately 200 m from the experimental vineyard. This station is therefore considered representative of the local climatic conditions experienced by the vines during both growing seasons. Although no in-row or plant-level microclimatic sensors were deployed, the proximity of the station ensures a reliable characterization of air temperature and precipitation at the field scale.

2.4. Yield and Yield Components

Yield and yield components were assessed at harvest in each growing season. Harvest was carried out in late August 2023 and late August 2024, following the standard harvesting schedule of the vineyard. All treatments were harvested on the same dates to ensure comparability among treatments.

The following yield components were measured: number of bunches, total weight of bunches, weight of three representative bunches, length of bunches, and diameter of berries. These variables were selected to characterize yield formation and fruit morphology under contrasting irrigation regimes and biostimulant applications.

2.5. UAV Data Acquisition and Processing

Aerial surveys were conducted during two distinct phenological stages of the plants. The first set of flights occurred during the flowering period on 2 May 2023 and 25 May 2024, corresponding to BBCH 65. The second set of flights was conducted on 25 August 2023 and 27 August 2024, corresponding to the Berries Ripe for Harvest stage (BBCH 89) (Figure 4A,B;

Table 1. Flight mission details.

Flights	Date	Images Collected	Point Density (pt/cm2)	GSD (cm.pix−1)	Flying Altitude (m)
1	2 May 2023	1.074	0.23	1.04	23
2	25 August 2023	1.074	0.18	1.17	24
3	25 May 2024	1.032	0.18	1.15	24
4	27 August 2024	1.074	0.18	1.17	26

). Phenological stages were determined according to the classification system proposed by Lorenz et al. [30].

UAV surveys were conducted to acquire RGB, multispectral, and thermal imagery of the vineyard canopy, providing the raw data necessary to evaluate the relationship between remotely sensed information and agronomic performance. The collected datasets were processed using Agisoft Metashape Professional (version 2.1.1) [31] for image alignment and the generation of georeferenced orthomosaics and canopy temperature maps. Based on these outputs, a set of vegetation indices previously employed by Macedo et al. [9,32] was calculated, alongside a novel thermal-based indicator, the Simplified Crop Water Stress Index (CWSI_si) proposed by Bian et al. [33], to enhance the assessment of crop water status (Table 2). For interpretation purposes, CWSI_si values were classified into water stress levels based on threshold ranges commonly reported in the literature. According to DeJonge et al. [34], under mid-day conditions, values above 0.1 indicate the onset of plant water stress, values exceeding 0.5 correspond to moderate water stress, and values approaching 1.0 represent severe water stress.

The UAV surveys were performed using a DJI Matrice 210 RTK platform equipped with a MicaSense Altum sensor, which integrates high-resolution RGB, multispectral, and thermal imaging capabilities in a single payload. The multispectral sensor captures blue, green, red, red-edge, and near-infrared bands, while the thermal sensor enables direct retrieval of canopy surface temperature. All flights were conducted under clear-sky conditions, close to solar noon, to minimize illumination variability and shadow effects. Radiometric calibration procedures recommended by the manufacturer were applied, including the use of a calibrated reflectance panel, to ensure consistency across acquisition dates.

For each treatment and replicate, mean spectral and thermal values were extracted at the plot level by delineating canopy areas within the orthomosaics and averaging all canopy pixels, after removal of soil/background pixels. Pixel-level values were therefore aggregated by area rather than analyzed individually.

All spatial analyses and index (Table 2) calculations were performed using ArcGIS Pro 3.3.2 [35], ensuring accurate georeferencing and data extraction at the plot level. This workflow enabled the extraction of average spectral and thermal values for each treatment and replicate, facilitating a precise analysis of the vineyard’s spatial variability.

Table 2. Vegetation indices were used.

Index	Formula	Reference	Equation Number
Green Leaf Index	$GLI={\displaystyle \frac{(2\times Green-Red-Blue)}{(2\times Green+Red+Blue)}}$	[36]	(1)
Green Normalized Vegetation Index	$GNDVI={\displaystyle \frac{(Nir-Green)}{(Nir+Green)}}$	[37]	(2)
Normalized Difference Red Edge	$NDRE={\displaystyle \frac{(Nir-Re)}{(Nir+Re)}}$	[38]	(3)
Normalized Difference Vegetation Index	$NDVI={\displaystyle \frac{(Nir-Red)}{(Nir+Red)}}$	[39]	(4)
Normalized Green Red Difference Index	$NGRDI={\displaystyle \frac{(Green-Red)}{(Green+Red)}}$	[40]	(5)
Crop Water Stress Index Simplified	${CWSI}_{si}={\displaystyle \frac{T_c-T_{wet}}{T_{dry}-T_{wet}}}$	[33]	(6)

Note: Blue—blue reflectance; Green—green reflectance; Red—red reflectance; Re—red edge reflectance; Nir—near-infrared reflectance. T_c is the average canopy temperature acquired using the UAV thermal images after removal of soil pixels, T_wet is the mean of the lowest 0.5% of canopy temperatures, and T_dry is the mean of the highest 0.5% of canopy temperatures.

Table 2. Vegetation indices were used.

Index	Formula	Reference	Equation Number
Green Leaf Index	$GLI={\displaystyle \frac{(2\times Green-Red-Blue)}{(2\times Green+Red+Blue)}}$	[36]	(1)
Green Normalized Vegetation Index	$GNDVI={\displaystyle \frac{(Nir-Green)}{(Nir+Green)}}$	[37]	(2)
Normalized Difference Red Edge	$NDRE={\displaystyle \frac{(Nir-Re)}{(Nir+Re)}}$	[38]	(3)
Normalized Difference Vegetation Index	$NDVI={\displaystyle \frac{(Nir-Red)}{(Nir+Red)}}$	[39]	(4)
Normalized Green Red Difference Index	$NGRDI={\displaystyle \frac{(Green-Red)}{(Green+Red)}}$	[40]	(5)
Crop Water Stress Index Simplified	${CWSI}_{si}={\displaystyle \frac{T_c-T_{wet}}{T_{dry}-T_{wet}}}$	[33]	(6)

Note: Blue—blue reflectance; Green—green reflectance; Red—red reflectance; Re—red edge reflectance; Nir—near-infrared reflectance. T_c is the average canopy temperature acquired using the UAV thermal images after removal of soil pixels, T_wet is the mean of the lowest 0.5% of canopy temperatures, and T_dry is the mean of the highest 0.5% of canopy temperatures.

2.6. Statistical Analysis

Data analysis was conducted using the Jamovi software (version 2.6) [41]. Before conducting inferential analyses, the distribution of the data was assessed using the Shapiro–Wilk test to evaluate the assumption of normality of the data. The results revealed a significant deviation from normality at the 1% level, indicating that the data were not normally distributed. Spearman’s rank correlation was applied to assess the monotonic relationships between variables related to grapevine production. The strengths of the correlation coefficients were interpreted according to the guidelines proposed by Schober et al. [42], with values classified as very weak (0.00–0.10), weak (0.10–0.39), moderate (0.40–0.69), strong (0.70–0.89), and very strong (0.90–1.00).

3. Results

3.1. Agronomic Performance Under Different Treatments (2023)

The combined effects of irrigation regime and amino-acid foliar application influenced grapevine productivity during the 2023 growing season (

Table 3. Effect of amino acids on agronomic parameters of grapevine crops in 2023.

Treatments	Number of Bunches	Total Bunch Weight (kg Plant−1)	Weight of 3 Bunches (g)	Length of Bunches (cm)	Diameter of Berries (mm)
T1	9.13 ± 4.34 b	1888.54 ± 1318 b	246.74 ± 60.82 a	15.22 ± 1.72 a	17.11 ± 1.28 a
T2	12.26 ± 8.24 b	2714.71 ± 2040 b	251.08 ± 124.89 a	13.98 ± 4.33 a	16.04 ± 4.47 a
T3	10.33 ± 7.69 b	1976.93 ± 1794 b	237.50 ± 104.97 a	14.59 ± 4.33 a	16.46 ± 1.53 a
T4	19.46 ± 10.04 a	3867.96 ± 1736 a	253.33 ± 54.85 a	15.10 ± 1.15 a	17.43 ± 1.41 a

Note: Means followed by the same letter do not differ by the Scott-Knott test at a 5% probability level of error. Average ± standard deviation.

), with amino acids tested under non-irrigated conditions (T1–T3) and compared against a fully irrigated control (T4). Treatment T4, representing conventional irrigation without amino acid application, exhibited superior agronomic performance, with a statistically higher number of bunches (19.46 ± 10.04) and total bunch weight (3867.96 ± 1736 kg) than the other treatments. These results reinforce the critical role of an adequate water supply in maximizing grapevine yield under Macaronesian conditions.

Among the non-irrigated treatments, no significant differences were detected for any agronomic parameter in 2023. Nevertheless, T2 consistently showed higher mean values for yield components than the untreated control (T1), including number of bunches and total bunch weight. These differences represent numerical trends rather than statistically supported effects and warrant confirmation in multi-year trials with increased replication and statistical power.

Treatment T3 did not produce noticeable improvements over T1, indicating the limited efficacy of this amino acid (pipecolic acid) in the environmental and agronomic context of the 2023 season. Although some increases were observed in bunch number and weight (10.33 ± 7.69 and 1976.93 ± 1794 kg, respectively), these differences were not significant.

No significant treatment effects were observed for the weight of three representative bunches, bunch length, or berry diameter, indicating that these morphometric traits were less sensitive to irrigation status and biostimulant application during this cycle. This suggests that the primary response to the treatments was quantitative (yield components) rather than qualitative (fruit morphology).

Overall, the results from the 2023 season highlight the importance of water availability for grapevine productivity, and the potential effects of pyroglutamic acid application under water-limited conditions.

3.2. Agronomic Performance Under Different Treatments (2024)

In contrast to the 2023 results, no statistically significant differences were observed among the treatments for any agronomic parameter evaluated in 2024 (

Table 4. Effect of amino acids on agronomic parameters of grapevine crops in 2024.

Treatments	Number of Bunches	Total Bunch Weight (kg Plant−1)	Weight of 3 Bunches (g)	Length of Bunches (cm)	Diameter of Berries (mm)
T1	2.46 ± 1.49 a	0.34 ± 0.31 a	205.38 ± 104.41 a	16.40 ± 3.62 a	13.93 ± 1.19 a
T2	4.18 ± 2.55 a	0.65 ± 0.44 a	166.18 ± 73.10 a	14.90 ± 3.65 a	13.54 ± 1.31 a
T3	3.45 ± 2.90 a	0.69 ± 0.92 a	167.75 ± 93.65 a	14.40 ± 4.05 a	13.34 ± 1.00 a
T4	3.30 ± 3.00 a	0.53 ± 0.76 a	148.26 ± 60.32 a	14.29 ± 2.93 a	13.62 ± 1.24 a

Note: Means followed by the same letter do not differ by the Scott-Knott test at a 5% probability level of error. Average ± standard deviation.

). This general lack of statistical separation is likely attributable to more severe climatic conditions, particularly the persistence of very low rainfall from flowering through early berry development (approximately May–June), followed by continued precipitation scarcity until harvest, as shown in Figure 3. These prolonged drought conditions likely constrained grapevine development across all treatments, thereby limiting the potential for differential treatment effects.

Although differences were not significant, numerical variation among treatments was observed. T2 recorded the highest number of bunches per plant 4.18 ± 2.55 and a total bunch weight 0.65 ± 0.44 kg, while T3 showed the highest total weight 0.69 ± 0.92 kg. The irrigated treatment (T4) yielded 3.30 ± 3.00 for the number of bunches and 0.53 ± 0.76 kg for the Total Bunches Weight. T1 presented the lowest number of bunches (2.46 ± 1.49) and total weight (0.34 ± 0.31) kg.

Morphometric traits, including average bunch weight, bunch length, and berry diameter, showed similar values across treatments, with no statistical differences detected. Overall productivity in 2024 was markedly lower than in 2023 across all treatments, although amino acid treatments (T2 and T3) showed numerically higher values than the untreated control in some yield-related variables.

3.3. Interannual Comparison of Canopy Water Stress After Biostimulant Application (2023 vs. 2024)

Figure 5 presents thermal maps obtained via UAV based on the Simplified Crop Water Stress Index (CWSI_si) for 2023 (top image) and 2024 (bottom image), both captured shortly after the application of biostimulants during the reproductive stage of the grapevine. CWSI_si values were classified into ten intervals representing increasing levels of thermal stress, with a color gradient ranging from purple (low stress) to dark red (severe stress).

Climatic analysis for the January–May period (Figure 2 and Figure 3) revealed that 2023 experienced slightly higher average temperatures and a more pronounced accumulated water deficit. Despite these conditions, grapevines showed an overall more favorable thermal response in 2023 than in 2024, suggesting that the temporal distribution and intensity of stress events in 2024 may have constrained vine physiological performance, even under less extreme average climatic conditions.

In both years, the T1 exhibited elevated canopy water stress, with a wide spatial distribution of high CWSI_si values. In 2023, T1 presented a mean CWSI_si of 0.443, while in 2024 the mean decreased to 0.341, reflecting differences in the timing and spatial distribution of water stress rather than an overall improvement in environmental conditions (

Table 5. CWSI_si values (replicates and means) for each treatment shortly after biostimulant application in 2023 and 2024.

	2 May 2023				25 May 2024
	r1	r2	r3	Mean	r1	r2	r3	Mean
T1	0.38370	0.46600	0.47899	0.44290	0.43954	0.25195	0.33194	0.34114
T2	0.47738	0.40341	0.42895	0.43658	0.36494	0.26999	0.43207	0.35567
T3	0.41018	0.42136	0.49511	0.4422	0.29659	0.36278	0.24048	0.29995
T4	0.39036	0.41937	0.48459	0.43144	0.35667	0.25091	0.48446	0.36401

Note: T1—Without irrigation and amino acids; T2—Pyroglutamic acid; T3—Pipecolic acid; T4—Conventional irrigation; r1, r2, r3—replicates.

). The dominance of warm colors in both seasons confirms the strong dependence of canopy thermal status on water availability and physiological regulation.

Treatment T2 showed consistently lower canopy water stress than the untreated control in both years. In 2023, T2 exhibited a mean CWSI_si of 0.437, slightly lower than T1 and comparable to the irrigated treatment, while in 2024, the mean value further decreased to 0.356. Spatially, T2 was characterized by a predominance of cooler tones (purple to cyan), indicating effective short-term thermal regulation following application. Although not always presenting the lowest absolute values, the stability of its response across contrasting climatic conditions highlights a rapid and reproducible metabolic and physiological adjustment induced by the biostimulant.

Treatment T3 exhibited an intermediate but variable response between years. In 2023, T3 showed the highest mean CWSI_si value (0.442), comparable to the untreated control, indicating limited effectiveness under drier early-season conditions. In contrast, in 2024, T3 presented the lowest mean CWSI_si among all treatments (0.300), suggesting a stronger response under conditions where stress intensity and timing were more favorable at the moment of application. This interannual variability, also evident in the heterogeneous spatial patterns observed in the thermal maps, suggests that the physiological response induced by this treatment may be more dependent on environmental context and less stable across seasons.

The T4 served as the physiological reference. In 2023, T4 showed a relatively low mean CWSI_si (0.431), indicating a relatively stable canopy water status shortly after irrigation. In 2024, the mean value (0.364) remained low but did not differ markedly from the best-performing non-irrigated treatments. This behavior indicates that, at this phenological stage, irrigation ensured stable plant water status but did not always result in distinctly lower canopy temperatures than those achieved through effective biostimulant-induced physiological regulation.

Overall, the integration of spatial thermal patterns with mean CWSI_si values indicates that biostimulant treatments were capable of partially mitigating canopy water stress shortly after application, with treatment-specific and year-dependent responses. While T2 showed the most consistent interannual behavior, T3 exhibited the strongest reduction in CWSI_si under favorable climatic conditions. These results highlight that biostimulant efficiency is closely linked to environmental context and underscores their potential role as complementary tools to irrigation for vineyard water stress management during the critical reproductive stage.

Although 2024 had a stronger precipitation deficit, CWSI_si values remained within ranges comparable to those observed in 2023, reflecting differences in within-flight thermal distributions rather than a direct measure of seasonal rainfall.

Table 5 and

Table 6. Mean Simplified Crop Water Stress Index (CWSI_si) values at harvest for each treatment in 2023 and 2024.

	25 August 2023				27 August 2024
	r1	r2	r3	Mean	r1	r2	r3	Mean
T1	0.41255	0.33324	0.36559	0.37046	0.392105	0.176471	0.364664	0.31108
T2	0.31953	0.36482	0.38205	0.35547	0.339801	0.327399	0.249414	0.30543
T3	0.38542	0.34868	0.43926	0.39112	0.39200	0.37541	0.47167	0.41303
T4	0.28785	0.31929	0.38155	0.32957	0.29585	0.46229	0.30937	0.35584

Note: T1—Without irrigation and amino acids; T2—Pyroglutamic acid; T3—Pipecolic acid; T4—Conventional irrigation; r1, r2, r3—replicates.

present CWSI_si values measured at two key time points: immediately before the application of amino acid-based biostimulants and at harvest. Values are reported per treatment and replicate (n = 3), together with treatment means, to capture both within-treatment variability and temporal changes in canopy water status. Given the limited number of replicates and the repeated-measures nature of the observations, these data are presented descriptively to support temporal and comparative interpretation rather than formal post hoc statistical separation.

3.4. Interannual Comparison of Canopy Water Stress at Harvest in 2023 and 2024

Figure 6 shows the thermal maps obtained by the UAV during the grape harvests of 2023 (top image) and 2024 (bottom image), enabling a spatial assessment of canopy water stress at the end of the production cycle. The images were processed based on the Simplified Crop Water Stress Index (CWSI_si) and categorized into ten intervals representing increasing levels of thermal stress, with a color scale ranging from purple (minimum stress) to dark red (maximum stress).

The climatological analysis for the May–August period (Figure 2 and Figure 3) indicated that 2023 was characterized by higher average temperatures and substantially lower rainfall, resulting in a drier environment prone to intensified water stress. In contrast, 2024 exhibited differences in rainfall distribution and thermal patterns during the same period, which resulted in distinct stress dynamics compared with 2023. These interannual differences were clearly reflected in both the spatial distribution and the mean CWSI_si values observed at harvest.

Plants in T1 exhibited high levels of canopy water stress in both seasons, with a predominance of warm colors (yellow, orange, and red) concentrated in the upper CWSI_si classes. At harvest, thermal maps revealed that T1 exhibited a higher spatial coverage of the upper CWSI_si classes (e.g., >0.65), indicating widespread canopy water stress. However, mean CWSI_si values calculated at the plot level remained within the ranges reported in Table 6. While these high-stress classes were more spatially extensive in 2023, mean plot-level CWSI_si values were lower and less contrasted in 2024, reflecting differences in stress distribution rather than absolute canopy temperature.

Plants in T2 demonstrated a clear reduction in canopy water stress compared with the untreated control. In 2023, thermal maps indicated that T2 was mainly associated with intermediate CWSI_si classes, while the mean plot-level CWSI_si value was 0.355, indicating moderate stress mitigation (Table 6). In 2024, under different environmental conditions at the time of acquisition, the spatial distribution shifted towards cooler classes (0.265–0.686), and the mean CWSI_si further decreased to 0.305, the lowest among the non-irrigated treatments. This interannual consistency confirms the capacity of this biostimulant to sustain improved leaf water status and thermal regulation until the end of the production cycle.

The T3 resulted in a more unstable and heterogeneous thermal response. Thermal maps revealed the presence of high-stress CWSI_si classes (>0.65) within T3, indicating pronounced spatial heterogeneity. However, mean plot-level CWSI_si values remained within the ranges reported in Table 6. This behavior suggests that, although metabolic signaling may have been activated, the induced physiological responses were insufficiently uniform or effective to ensure reliable stress mitigation under field conditions.

Treatment T4, representing conventional irrigation, was used as a physiological reference. In 2023, thermal maps showed that T4 was predominantly associated with lower and intermediate CWSI_si classes, while the mean plot-level CWSI_si value was 0.330, indicating a relatively stable canopy water status under irrigation (Table 6). In 2024, although the spatial distribution remained relatively stable (0.355–0.686), the mean CWSI_si increased to 0.356, slightly exceeding that observed in T2. This response may be associated with differences in irrigation efficiency, canopy vigor, or atmospheric demand during the ripening period.

Overall, the integration of spatial thermal patterns with mean CWSI_si values at harvest confirms that T2 was capable of consistently promoting a canopy thermal response under non-irrigated conditions that approached that observed under irrigation. Its ability to reduce water stress across two climatically contrasting seasons reinforces its practical applicability as a sustainable strategy for vineyard management under conditions of increasing water scarcity.

3.5. Correlation Between Morphological Parameters and Vegetation Indices in 2023

Correlation analysis revealed strong and statistically significant relationships among yield components across treatments (

Table 7. Significant Spearman correlations (ρ) between key morphological, yield, and spectral/thermal parameters in the 2023 season.

Relationship	Global	T1	T2	T3	T4
Total weight of bunches × Number of bunches	0.884 ***	0.643 **	0.922 ***	0.856 ***	0.874 ***
Total weight of bunches × Weight of 3 bunches	0.505 ***	–	0.821 ***	0.661 **	–
Total weight of bunches × Length of bunches	0.354 **	–	–	0.646 *	–
Total weight of bunches × Diameter of berries	0.292 *	–	–	-	–
Length of bunches × Weight of 3 bunches	0.640 ***	0.630 *	0.525 *	0.825 ***	0.698 **
Diameter of berries × Weight of 3 bunches	0.441 ***	0.604 *	–	–	–
Diameter of berries × Length of bunches	0.352 **	–	0.539 *	–	–
Number of bunches × Weight of 3 bunches	–	–	0.606 *	–	−0.541 *
Length of bunches × Number of bunches	–	–	–	–	−0.521 *
Number of bunches × CWSIsi	–	–	0.579 *	–	0.632 *
Length of bunches × GLI	–	–	−0.550 *	–	–
CWSIsi × GLI	−0.820 ***	−0.771 **	−0.682 **	−0.811 ***	−0.904 ***
CWSIsi × GNDVI	−0.588 ***	−0.832 ***	−0.686 **	−0.889 ***	−0.893 ***
CWSIsi × NDRE	−0.587 ***	−0.836 ***	−0.721 **	−0.889 ***	−0.829 ***
CWSIsi × NDVI	−0.776 ***	−0.850 ***	−0.711 **	−0.918 ***	−0.900 ***
CWSIsi × NGRDI	−0.823 ***	−0.821 ***	−0.639 *	−0.846 ***	−0.932 ***

Note: correlation is significant when: * p < 0.05, ** p < 0.01, *** p < 0.001; “–” indicates non-significant relationships.

). Globally, total bunch weight was strongly and positively correlated with the number of bunches (ρ = 0.884 ***), indicating that yield formation was primarily driven by bunch number. Moderate but significant correlations were also observed between total bunch weight and the weight of three representative bunches (ρ = 0.505 ***), bunch length (ρ = 0.354 **), and berry diameter (ρ = 0.292 *), reflecting the contribution of cluster and berry size to total yield.

Treatment-specific correlation patterns differed markedly among treatments. In the T1, the strongest relationship was observed between total bunch weight and number of bunches (ρ = 0.643 **), while correlations involving bunch size variables, such as weight of three representative bunches and bunch length, were weak or not significant. This indicates a yield structure mainly driven by bunch numbers under unmanaged water stress conditions. In T2, the highest correlations were detected between total bunch weight and number of bunches (ρ = 0.922 ***) and between total bunch weight and the weight of three representative bunches (ρ = 0.821 ***), suggesting that both bunch number and bunch mass contributed substantially to yield formation. Additionally, a significant positive correlation between bunch length and weight of three representative bunches (ρ = 0.525 ***) indicates a coherent relationship between cluster morphology and cluster mass under this treatment.

Treatment T3 showed its strongest correlation between bunch length and weight of three representative bunches (ρ = 0.825 ***), highlighting the relevance of cluster size attributes in determining bunch mass under this treatment, despite fewer significant relationships among other yield components.

Under T4 treatment, a very strong correlation was observed between total bunch weight and number of bunches (ρ = 0.874 ***), whereas relationships involving bunch size variables were less consistently significant, reflecting a yield structure primarily driven by reproductive load under irrigated conditions.

Across all treatments and at the global level, the Simplified Crop Water Stress Index (CWSI_si) was consistently and negatively correlated with all vegetation indices analyzed, including GLI, GNDVI, NDRE, NDVI, and NGRDI. Global correlation coefficients ranged from −0.588 to −0.823 (p < 0.001), indicating that increasing thermal-based water stress was associated with reduced canopy vigor as captured by spectral indices. Similar patterns were observed within individual treatments, with particularly strong negative correlations in T2, T3, and T4 (|r| > 0.68, p < 0.01).

3.6. Correlation Between Morphological Parameters and Vegetation Indices in 2024

In the 2024 growing season, strong and statistically significant correlations were observed among yield components, confirming coherent relationships among yield components despite increased water stress conditions (

Table 8. Significant Spearman correlations (ρ) between key morphological, yield, and spectral/thermal parameters in the 2024 season.

Relationship	Global	T1	T2	T3	T4
Total weight of bunches × Number of bunches	0.844 ***	0.671 *	0.919 ***	0.901 ***	0.778 **
Total weight of bunches × Weight of 3 bunches	0.709 ***	0.819 **	–	0.618 *	0.890 ***
Total weight of bunches × Length of bunches	0.629 ***	0.718 **	–	–	0.871 ***
Total weight of bunches × Diameter of berries	0.500 ***	–	0.682 *	0.665 *	–
Number of bunches × Weight of 3 bunches	0.462 ***	0.614 *	–	–	–
Number of bunches × Length of bunches	0.488 ***	0.629 *	–	–	0.599 *
Number of bunches × Diameter of berries	0.460 **	–	–	0.749 **	–
Weight of 3 bunches × Length of bunches	0.850 ***	0.883 ***	0.756 **	0.645 *	0.835 ***
Weight of 3 bunches × Diameter of berries	0.543 ***	0.747 **	0.627 *	–	–
Diameter of berries × Length of bunches	0.370 **	–	–	–	–
Total weight of bunches × CWSIsi	–	–	0.709 *	–	−0.610 *
Number of bunches × CWSIsi	–	–	0.711 *	–	–
Total weight of bunches × NGRDI	0.456 **	–	–	–	0.665 *
Total weight of bunches × NDVI	0.332 *	–	–	–	–
Total weight of bunches × GLI	0.437 **	–	–	–	–
Number of bunches × NGRDI	0.293 *	–	–	–	–
Weight of 3 bunches × NGRDI	0.293 *	–	–	–	–
Diameter of berries × NGRDI	0.289 *	–	–	–	–
Diameter of berries × NDVI	0.344 *	–	–	–	–
CWSIsi × GNDVI	−0.492 ***	–	–	–	−0.676 *
CWSIsi × NDRE	−0.618 ***	−0.610 *	–	–	−0.835 ***
CWSIsi × NDVI	−0.465 ***	−0.577 *	–	–	−0.857 ***
CWSIsi × GLI	–	–	–	–	−0.610 *

Note: correlation is significant when: * p < 0.05, ** p < 0.01, *** p < 0.001; “–” indicates non-significant relationships.

). At the global level, total bunch weight was strongly correlated with the number of bunches (ρ = 0.844 ***), as well as with the weight of three representative bunches (ρ = 0.709 ***) and bunch length (ρ = 0.629 ***). These relationships indicate that, even under severe drought conditions, yield variability was primarily driven by reproductive load and cluster size attributes.

In the T1, the strongest correlations were observed between the weight of three representative bunches and bunch length (ρ = 0.883 ***), followed by relationships between total bunch weight and weight of three bunches (ρ = 0.819 **), and between total bunch weight and bunch length (ρ = 0.718 **). These results indicate that, under unmanaged water stress, yield formation in 2024 relied more strongly on cluster morphology and bunch mass than on bunch number alone.

In the pyroglutamic acid treatment (T2), the highest correlation was detected between total bunch weight and number of bunches (ρ = 0.919 ***), highlighting the dominant role of bunch number in yield determination under this treatment. Additionally, significant positive correlations were observed between the weight of three representative bunches and bunch length (ρ = 0.756 **), and between the weight of three bunches and berry diameter (ρ = 0.627 *), suggesting a coordinated relationship between cluster size and berry development.

The pipecolic acid treatment (T3) exhibited strong correlations between total bunch weight and number of bunches (ρ = 0.901 ***), as well as between number of bunches and berry diameter (ρ = 0.749 **). However, fewer significant correlations were observed among other yield components, indicating a more selective coupling between reproductive load and berry size under this treatment in 2024.

Under conventional irrigation management (T4), very strong correlations were observed between total bunch weight and number of bunches (ρ = 0.778 **), weight of three representative bunches (ρ = 0.890 ***), and bunch length (ρ = 0.871 ***). This pattern reflects a highly coherent yield structure under irrigated conditions, with both bunch number and cluster size contributing substantially to yield formation.

In 2024, significant correlations between yield components and spectral or thermal indicators were more limited than in the previous season. Nevertheless, treatment-specific relationships involving CWSI_si were detected. In the pyroglutamic acid treatment (T2), total bunch weight and number of bunches were positively correlated with CWSI_si (ρ = 0.709 * and ρ = 0.711 *, respectively; p < 0.05), whereas under conventional irrigation (T4), total bunch weight was negatively correlated with CWSI_si (ρ = −0.610 *). This positive association suggests that, under severe drought, higher-yielding vines within this treatment may have experienced greater transpirational demand and canopy temperature, rather than reduced physiological performance.

Across treatments, CWSI_si showed consistent and significant negative correlations with vegetation indices, particularly NDVI and NDRE. At the global level, CWSI_si was negatively correlated with NDVI (ρ = −0.465 *, p < 0.001) and NDRE (ρ = −0.618 ***). Similar patterns were observed within T1 and T4, where stronger negative correlations were detected (p > 0.57, p < 0.05), indicating that increased thermal-based water stress was associated with reduced canopy vigor under severe drought conditions.

4. Discussion

4.1. Agronomic Performance Under Contrasting Climatic Conditions

The interpretation of treatment effects should be framed within the context of strong interannual variability typical of perennial vine systems.

The agronomic responses of grapevines to the application of amino acid-based biostimulants and varying irrigation regimes were evaluated across two consecutive growing seasons with distinct climatic profiles: 2023, characterized by moderate drought, and 2024, marked by severe water deficit. A comparative analysis of both years provided insights into the efficacy and limitations of biostimulant treatments under different stress intensities.

In 2023, significant differences were observed among the treatments for both the number of bunches and total bunch weight, with T4 showing the highest productivity (Table 3). These findings reinforce the well-established importance of adequate watering to maximize grapevine yields, particularly in Mediterranean climates [2,7].

Importantly, these differences were not statistically significant at the 95% confidence level, likely due to high within-treatment variability typical of field-grown perennial crops. Therefore, the observed patterns for T2 should be interpreted as indicative trends rather than conclusive evidence of treatment efficacy.

However, among the water-stressed treatments, T2 consistently exhibited higher mean values for the number and weight of bunches compared to T1, although these differences were not statistically significant. These numerical trends are consistent with previous reports on pyroglutamic acid under drought conditions [20], but in the present study they should be interpreted as indicative physiological responses rather than confirmed treatment effects. The observed patterns therefore suggest potential modulation of vine responses under moderate stress, which requires confirmation through additional seasons and higher statistical power.

The large standard deviations observed in agronomic variables, particularly total bunch weight, are consistent with the intrinsic heterogeneity of vineyard field conditions and perennial vine systems. Differences in vine vigor, rooting depth, canopy architecture, and fine-scale soil and microclimatic variability along vineyard rows can generate substantial within-treatment variation, especially under drought conditions. This level of variability reduces statistical power and can mask moderate treatment effects unless larger plot numbers and/or additional seasons are included. This field heterogeneity is reflected in the high coefficients of variation and large standard deviations reported for several agronomic variables in Table 3 and Table 4, which reduces statistical power and increases the likelihood of non-significant treatment separation.

In 2024, the persistence of very low rainfall from flowering through early berry development and the continued precipitation deficit until harvest (Figure 3) resulted in overall low productivity and the absence of statistically significant differences among treatments (Table 4). More extreme drought conditions likely masked the treatment effects by uniformly constraining grapevine development, thereby diminishing inter-treatment variability and limiting the detection of treatment-specific responses.

The absence of statistical differences among treatments in 2024 and the low overall productivity indicate that severe drought constrained vine performance across the entire experiment. Although T4 represented conventional irrigation, it followed an empirical grower practice without sensor-based scheduling or quantified irrigation amounts, which may not have been sufficient to fully offset extreme atmospheric demand. Under such conditions, irrigated vines may not necessarily outperform non-irrigated treatments, and the numerically higher values observed for T2 and T3 should be interpreted cautiously as field variability rather than confirmed yield gains. Nevertheless, T2 again displayed the highest number of bunches (4.18 ± 2.55) and a relatively high total weight (0.65 ± 0.44 kg), representing a numerical trend that may indicate a transient physiological adjustment, rather than a confirmed resilience mechanism under extreme drought conditions. Notably, even the T4 exhibited reduced productivity in 2024, further illustrating the severity of climatic constraints and the limited buffering capacity of irrigation alone under such extreme conditions.

Interestingly, T3 recorded the highest total weight in 2024 (0.69 ± 0.92 kg), despite considerable variability. Although T3 showed limited efficacy in 2023, its relatively high total weight in 2024, despite high variability, represents a numerically higher but highly variable response under extreme drought conditions, which cannot be interpreted as a consistent treatment effect, possibly linked to the antioxidant-inducing role of pipecolic acid, as reported by Wang et al. [22] for tomatoes under drought conditions. This variability highlights the need for further investigation of the dosage, timing, and crop-specific responses.

In both seasons, morphological parameters, such as the weight of the three bunches, bunch length, and berry diameter, did not differ significantly among the treatments. This suggests that these traits may be more genetically stable and less influenced by short-term agronomic interventions, a notion supported by Keller [24], who emphasized the structural resilience of perennial crops, such as grapevines, to episodic stress events.

Taken together, this two-year study highlights the strong interaction between environmental conditions and treatment responses. While irrigation remains the most effective strategy under moderate stress, amino acid-based biostimulants, particularly pyroglutamic acid, exhibited consistent numerical and physiological trends indicative of potential complementary roles in supporting vineyard performance under water-limited conditions, although these effects were not statistically confirmed. Their performance under extreme stress appears to be constrained, underscoring the importance of integrated management strategies that combine biostimulants with irrigation optimization and stress monitoring.

4.2. Integrated Discussion: Spectral–Morphological Relationships Across Two Drought-Affected Seasons

Correlation analyses conducted over two consecutive grapevine growing seasons (2023 and 2024) under water deficit conditions provided consistent evidence of strong associations between key agronomic traits and vegetation indices, reinforcing the potential of remote sensing tools to support yield estimation and stress diagnosis in grapevines.

Although correlation analyses are useful to reveal consistent associations among spectral, thermal, and agronomic variables, they do not establish causality. The observed negative relationships between CWSI_si and vegetation indices are physiologically plausible because water deficit tends to reduce stomatal conductance and transpiration cooling, raising canopy temperature, while simultaneously constraining canopy development and pigment status, which lowers reflectance-based vigor indices. However, these processes can also be influenced by confounding factors such as canopy geometry, soil exposure, and within-row heterogeneity. Therefore, the correlations reported here should be interpreted as exploratory evidence supporting mechanistic consistency, rather than as proof of direct cause–effect relationships.

In both years, a strong positive correlation was observed between the number of bunches and the total weight of the bunches (0.884 *** in 2023 and 0.844 *** in 2024), confirming that fruit set is a major determinant of total yield. These findings are aligned with core viticultural principles and are supported by studies by Intrigliolo & Castel [43] and Keller [24], which highlight the centrality of reproductive parameters, particularly fruit set and cluster development, in yield formation. Image-based modelling was used previously by authors [44,45] to show that visible morphological traits, such as bunch volume and berry count, are highly predictive of bunch weight, underscoring the relationship between reproductive development and yield outcomes. Similarly, it was demonstrated that grape bunch volume and berry number extracted from color images using deep learning are reliable indicators for yield estimation in vineyards [46].

Additional correlations observed in both seasons revealed significant interrelationships between bunch weight, length, and berry size, particularly between the weight of the three bunches and bunch length (0.640 *** in 2023; 0.850 *** in 2024) and between the weight of the three bunches and berry diameter (0.441 *** in 2023; 0.543 *** in 2024). Furthermore, bunch length and berry diameter showed moderate but consistent associations (0.352 ** in 2023; 0.370 ** in 2024), indicating that both parameters contribute to fruit mass and, ultimately, to total productivity.

These associations reflect the complex but coherent structure of yield components and are aligned with the current understanding of viticulture, which emphasizes bunch architecture and berry development as key indicators of crop performance [45,47]. These patterns suggest that larger and longer bunches and larger berries contribute meaningfully to productivity, which is consistent with the findings of Palacios et al. [47], who emphasized the cumulative role of berry size and bunch architecture in the final yield potential of grapevines using image-based prediction models.

4.3. Interannual Comparison of Canopy Water Stress Following Biostimulant Application (2023 vs. 2024)

The simplified CWSI approach used here (CWSI_si) is derived from canopy temperature distributions within each flight, using T_wet and T_dry defined from the lowest and highest canopy temperature percentiles of the same dataset. As a result, CWSI_si is a relative indicator of canopy thermal status at the time of image acquisition, and its values are not expected to scale linearly with seasonal precipitation deficits. Under extreme drought, physiological responses such as stomatal closure and reduced transpiration can also lead to partial saturation of canopy temperature signals, limiting further increases in thermal stress indices even when soil water availability remains very low.

The use of UAV-derived thermal imagery enabled a spatially explicit analysis of canopy water stress across two viticultural seasons with contrasting climatic conditions in this study. In both the post-biostimulant application period and at harvest, the CWSI_si maps revealed consistent trends that underline the influence of both environmental and treatment variables. This approach is consistent with the findings of Atencia Payares et al. [48], who demonstrated that thermal imaging from UAVs accurately estimated the crop water status in Vitis vinifera cv. Merlot under semi-arid conditions, with a high correlation (r = 0.84) between the CWSI and stem water potential.

During the reproductive phase following biostimulant application, the years 2023 (moderate drought) and 2024 (severe drought) presented distinct thermal profiles. Treatment T2 consistently displayed the lowest CWSI_si values, indicating superior leaf cooling capacity and physiological activity under water-limited conditions. This was reflected in the strong negative correlations between CWSI_si and spectral indices in both years, such as CWSI_si × NDVI = −0.918 *** in 2023 and CWSI_si × NDRE = −0.835 *** in 2024. These findings are consistent with those of Jiménez-Arias et al. [20], who demonstrated the efficacy of pyroglutamic acid in enhancing osmotic adjustment, photosynthetic capacity, and antioxidant responses in lettuce under drought stress. Although species-specific responses may vary, the interannual consistency of T2 observed in grapevines supports its broad physiological significance.

Moreover, the use of UAV-derived CWSI has been validated in viticulture as a reliable indicator of vine water status in the literature. Thermal imagery-based CWSI effectively discriminated between irrigated and non-irrigated vines in Vitis vinifera cv. Loureiro, which showed a strong correspondence with physiological indicators, such as stomatal conductance and leaf water potential [49].

At harvest, T2 continued to present thermal profiles comparable to those of the irrigated control (T4), with lower proportions of high-stress zones, even under severe drought in 2024. This sustained physiological advantage suggests that pyroglutamic acid may have prolonged protective effects. These results align with those of Batista Silva et al. [23], who demonstrated that amino acid metabolism plays a key role during stress recovery by acting in stress signaling, osmotic adjustment, redox regulation, and membrane stability, thereby extending plant resilience throughout the production cycle.

Treatment T3 exhibited good thermal-spectral performance in several cases, despite its higher intra-treatment variability. In 2023, CWSI_si × NGRDI reached −0.932 *** and CWSI_si × GLI reached −0.904 ***, indicating strong canopy cooling and high spectral vigor in specific areas. In 2024, T3 maintained robust negative correlations, such as CWSI_si × NDVI = −0.857 *** and CWSI_si × NDRE = −0.835 ***, suggesting a potential role for pipecolic acid under extreme stress, possibly linked to its function as a systemic acquired resistance (SAR) inducer and antioxidant modulator [25]. However, the high spatial heterogeneity and lack of consistently superior performance to T2 highlight the need for further investigation into dosage, application timing, and potential synergistic effects.

In contrast, T1 consistently exhibited the highest CWSI_si values, indicating elevated canopy temperatures and greater thermal stress. In 2024, this was further supported by a positive correlation between the number of bunches and CWSI_si (0.711 *), suggesting that a higher fruit load was associated with a reduced cooling capacity. These observations align with the findings of Burchard-Levine et al. [50], who reported that non-irrigated grapevines displayed significantly higher canopy temperatures and CWSI values that correlated negatively with leaf and stem water potential. Their study reinforced that the absence of effective mitigation strategies increases vine susceptibility to drought stress under Macaronesian conditions.

Overall, this interannual analysis confirmed that UAV-based thermal imaging combined with targeted biostimulant applications is a powerful approach for monitoring and managing vineyard water stress. T2 demonstrated repeatable and spatially coherent stress mitigation, validating its potential utility in rainfed viticulture. Nevertheless, future research should integrate physiological ground-truthing and multiyear trials to optimize biostimulant strategies under varying climatic scenarios.

4.4. Consistency and Divergence in Spectral Index Performance

In terms of vegetation indices, NDVI, NDRE, and GNDVI demonstrated very strong intercorrelations in both seasons, with coefficients consistently above 0.85 in treatment-level analyses, reflecting their shared sensitivity to chlorophyll content, canopy density, and plant vigor. Given this strong collinearity, the indices were not interpreted as independent predictors, but as complementary descriptors. NDVI primarily reflects overall canopy greenness and structure, NDRE tends to be more sensitive to chlorophyll variations in denser canopies, and GNDVI emphasizes changes in chlorophyll- and nitrogen-related vigor. In parallel, RGB-based indices (GLI and NGRDI) provide a low-cost proxy for visible greenness when multispectral sensors are not available. Thermal metrics (CWSI_si) capture a different physiological dimension linked to transpiration cooling and stomatal regulation, complementing reflectance-based indicators rather than duplicating them.

Recent studies have reinforced the applicability of these indices to viticulture. For instance, Giovos et al. [51] conducted a comprehensive review of over 90 vegetation indices used in viticulture, highlighting the effectiveness of the NDVI, NDRE, and GNDVI in assessing vine vigor and chlorophyll concentrations. Rehman et al. [52] compared the sensitivity of various vegetation indices and found that NDRE exhibited higher sensitivity than NDVI in detecting variations in canopy vigor, particularly in dense vegetation scenarios.

Similar patterns of strong intercorrelations among NDVI, NDRE, and GNDVI have been observed in other cropping systems. Macedo et al. [32] demonstrated that GNDVI and NDRE are more strongly correlated with yield and aboveground biomass than NDVI in corn fields on Madeira Island, highlighting their superior sensitivity to chlorophyll content and canopy vigor, which reinforces their applicability to diverse crops, including grapevines.

The GLI and NGRDI indices derived from the visible bands also showed strong correlations (0.985 *** in 2023 and approximately 0.96 *** in 2024), highlighting their value in RGB-based remote-sensing applications where multispectral imagery may not be available. Recent research has demonstrated the utility of RGB-based indices, such as GLI and NGRDI, in agricultural monitoring. For example, Panthakkan et al. [53] evaluated UAV-based RGB and multispectral vegetation indices in palm tree cultivation and concluded that RGB indices, such as GLI and NGRDI, provide comparable performance to multispectral indices in assessing vegetation health, offering a cost-effective alternative for large-scale agricultural monitoring.

Notably, the NGRDI was consistently correlated with yield-related traits, including the number of bunches (ρ = 0.293 *), total bunch weight (ρ = 0.456 **), and berry diameter (ρ = 0.289 *) in 2024, and with vegetation status in 2023, as evidenced by the strong negative correlations between CWSI_si and NGRDI (ρ = −0.823 *** global). These results suggest that NGRDI can serve as a low-cost RGB proxy for canopy vigor under drought, but its sensitivity to illumination/exposure and background effects requires cautious interpretation and further multi-season validation before being used as a standalone water-status indicator (Macedo et al., [32]; Matyukira et al., [54]).

The CWSI_si exhibited stronger and more consistent correlations in 2023 than in 2024. In 2023, the global correlations between CWSI_si and spectral indices ranged from −0.587 *** (NDRE) to −0.823 *** (NGRDI), with NDVI (−0.776 ***), GLI (−0.820 ***), and GNDVI (−0.588 ***), which also showed significant negative associations, indicating their effectiveness in identifying mild-to-moderate water stress through canopy temperature. In 2024, these correlations weakened in the global dataset, ranging from −0.465 *** (NDVI) to −0.618 *** (NDRE), with some indices, such as GLI and NGRDI, losing significance. This pattern suggests that under extreme drought conditions, thermal sensitivity may plateau, thereby reducing CWSI’s discriminative power [55].

These findings indicate that while thermal-based indicators are useful under moderate stress, their integration with reflectance-based indices (e.g., NDVI and NDRE) may offer a more resilient strategy for stress detection across various climatic scenarios, as supported by Matese & Di Gennaro [56] and Bellvert et al. [27]. The results across both years underline the robustness of the vegetation indices, especially NDVI, GNDVI, NDRE, and NGRDI, in correlating plant vigor and yield traits, even under contrasting drought conditions. These indices can serve as early warning systems for stress detection, allowing viticulturists to optimize irrigation, nutrient management, and harvest timing, thereby improving resource-use efficiency and sustainability in drought-prone viticultural systems [32,51,52,57]. Although the CWSI may have limited effectiveness under extreme stress conditions, reflectance-based indices retain their diagnostic value across both seasons. This reinforces the importance of integrated spectral monitoring, especially in Macaronesian environments that are increasingly affected by water scarcity.

The contrasting responses observed between pyroglutamic and pipecolic acid treatments may be partly explained by their distinct physiological roles in plant stress responses. Pyroglutamic acid has been shown to enhance drought tolerance in other species by supporting the maintenance of photosynthetic activity and stress resilience under water deficit [20]. This is consistent with the more stable spectral–thermal patterns observed under T2. In contrast, pipecolic acid functions as a key signaling molecule in plant stress physiology, involved in stress-induced metabolic reprogramming and defense responses [25]. Such signaling may contribute to more heterogeneous outcomes at the field scale, as reflected in the variable responses observed under T3.

A limitation of this field study is that, under high inter-plant variability and contrasting drought intensity, treatment effects on yield components may require larger replicated plot numbers and/or additional seasons to reach statistical separation. Moreover, future experimental designs could test amino-acid protectors under regulated deficit irrigation or partial irrigation maintenance after key phenological stages, to assess whether combined strategies provide more stable productivity gains than either approach alone.

Additionally, foliar applications were standardized operationally (same operator, equipment, concentration, and protocol), but the lack of a quantified spray volume per vine remains a limitation. This qualitative “full leaf wetting” criterion may introduce uncertainty in dose uniformity under field conditions and should be addressed in future trials through volumetric standardization.

In perennial crops such as grapevine, year-to-year differences in canopy architecture and vigor may arise from multiple interacting factors beyond water availability, including pruning intensity, shoot positioning, disease pressure, nutrient status, and fine-scale soil heterogeneity along vineyard rows. Even when management practices are intended to remain consistent across seasons, individual vines may express markedly different growth forms from one year to the next. This intrinsic interannual variability makes it challenging to fully isolate treatment effects, particularly under field conditions.

The use of UAV-based spatial monitoring partially addresses this complexity by enabling plot-scale assessment of canopy vigor and thermal patterns, thereby providing a spatial context for interpreting plant responses. However, while this approach strengthens the interpretation of treatment-induced trends, it cannot entirely disentangle management and plant-driven variability from water stress and biostimulant effects, reinforcing the importance of multi-year monitoring in perennial cropping systems.

Additional ground-based photographs across different phenological stages would enhance the visual interpretation of vine development and interannual variability and could be included as supplementary material in future work.

5. Conclusions

This study suggests that biostimulant application, particularly pyroglutamic acid (T2), may partially moderate the response to water stress, depending on stress intensity and phenological timing. Across two consecutive seasons (2023—moderate drought; 2024—severe drought), T2 showed consistentnumerical trends in productivity-related parameters and in canopy water status across seasons. In several instances, particularly at specific phenological stages and under moderate water stress, its performance in terms of canopy water stress, assessed via CWSI_si, was comparable to that of the conventional irrigation (T4), with consistently strong negative relationships between CWSI_si and vegetation indices, suggesting a relatively stable physiological response under varying abiotic stress conditions.

Agronomic evaluations revealed strong positive correlations between the number of bunches, total bunch weight, and morphological traits such as bunch length and berry diameter, reinforcing the contribution of reproductive structures to overall yield formation. The integration of these morphological parameters with spectral vegetation indices (NDVI, NDRE, GNDVI, and NGRDI) and UAV-derived thermal imagery enabled a multidimensional assessment of plant performance at both physiological and productive levels. In particular, the NGRDI demonstrated utility as a low-cost index, correlating with yield traits in 2024 and with vegetation status in 2023, while CWSI_si proved especially sensitive in detecting mild-to-moderate stress in 2023, although its discriminative power was reduced under extreme drought in 2024. This behavior indicates that biostimulant-induced thermal mitigation is strongly modulated by drought intensity, with reduced effectiveness under extreme stress conditions.

Thermal response patterns highlighted the stability of T2 under water-limited conditions, in contrast to the higher spatial and interannual variability observed for T3, which showed relevant but less consistent thermal–spectral responses across seasons. Although T3’s performance was less consistent across seasons, it showed notable peaks in thermal–spectral performance, particularly in 2023, when strong thermal–spectral coupling was observed. T1, which lacked both irrigation and biostimulant application, consistently presented the highest thermal stress values, confirming the limited physiological buffering capacity of vines under unmanaged stress conditions.

Therefore, pyroglutamic acid may represent a complementary management tool for viticulture in drought-prone regions, particularly as a complementary strategy to irrigation and precision monitoring, contributing to improved water-use efficiency and supporting sustainable productivity under moderate, and to a lesser extent, severe water stress. The results also highlight the usefulness of integrating UAV-based thermal imaging and vegetation indices into vineyard monitoring systems, as their combined use strengthens stress detection and yield prediction across contrasting climatic conditions. Nonetheless, further multi-year studies across diverse environmental contexts are necessary to validate these trends, optimize biostimulant application protocols, and better define their role as complementary tools within precision viticulture strategies under extreme drought scenarios. This study provides a baseline framework that can support continued seasonal monitoring of vineyard water stress and treatment responses in future assessments.