2.1. Plant Material, Experimental Layout, and Weather Course

In 2024, AV panel setting and vine assessment were performed in a small experimental vineyard at Residenza Gasparini, a facility of the Università Cattolica del Sacro Cuore located in the Piacenza campus (45.1°N, 9.6°E, Italy). The vineyard features four 50 m long, 35° NE–SW oriented rows planted with cvs. Ortrugo, MC, CS, and Barbera all grafted onto SO4 rootstock. The spacing is 2.5 × 1 m (between rows and in the row, respectively), with a resulting density of 4000 vines/ha. Vines are spur-pruned at a bud load of about 10–12 nodes per meter of row. The main cordon wire is located 90 cm from the ground, whereas single foliage wires are placed 30, 70, and 110 cm above the main wire to allow a maximum canopy height of about 2.2–2.4 m. Manual shoot trimming was performed twice (June 10 and July 16) in all varieties to limit maximum canopy height below 2.5 m.

The i-Pergola Agri PV Tech AV system (i-Pergola, Brescia, Italy) installed in the vineyard combines advanced hardware and software features for seamless integration with agricultural operations. The hardware setup includes 4 row sections, each measuring 16 m long, supported by 3 IPE 140 columns per row, spaced 8 m apart. The foundation depth is 2 m, and the above-ground height is 3 m, ensuring adequate clearance for vineyard machinery other than mechanical harvesting. The structure features a horizontal axis (100 × 100 × 3 mm section) and an east–west tracker with a ±90° movement range powered by a sleeve drive mechanism. Each row is equipped with 8 SunPower P6 495 COM panels (Sun Power Italia Srl, Milano, Italy), providing a total capacity of 16 kWh, complemented by a HUAWEI 15 kW inverter for efficient energy conversion.

Despite panels being mounted on the first part of all rows, readings were taken only on the two central rows planted with Malvasia MC and CS grapevines. Such a decision had the following motivation: (i) the two external rows acted as border rows; (ii) including all varieties would have prevented setting up a minimum number of chambers (i.e., 3) per treatment combination (4 cultivars × two exposures = 8). Real view of the AV panels mounted in the vineyard is reported in Figure S1.

For the whole duration of the trial (DOY 207 with the setup of solar panels to DOY 285 with the harvest of the AV treatment of CS), daily air temperature (minimum, mean, maximum) and precipitation were recorded by the Netsens weather station (Netsens Srl, Firenze, Italy) located adjacent to the vineyard.

2.2. Diurnal Total Light Interception

The amount of intercepted light in the presence/absence of the photovoltaic covers was measured according to the approach presented by [23, 24] with a few modifications. Light readings were taken via a scanner bar equipped with 64 low-cost phototransistors (BPW20RSilicon PN Photodiode, Vishay Telefunken, Heilbronn, Germany) sensing short-wave radiation in the 300–1100 nm waveband. The sensors were embedded in a lightweight aluminum bar (15 mm wide, 35 mm thick) and spaced at 35 mm to yield a total maximum measuring length of 2205 mm. Each sensor had a pixel area of 0.01225 m² and was covered with a Teflon layer to correct the readings in accordance with Lambert’s cosine law. The scanner bar was wired to an externally powered CR10 WP datalogger (Campbell Scientific Ltd, Lough-borough, UK), equipped with an AM 416 relay multiplexer. The data were then expressed as a photon flux density (PFD, μmol photons m⁻²s⁻¹), using the multiplying factor of 4.6 reported by [22]. Measurements of total canopy light interception (TCLI) were taken on August 13 (DOY 225) under perfectly clear-sky conditions and negligible wind speed at solar noon (about 2 p.m.) and 2 hours before and after solar noon along unchambered sections of the OF and AV test rows. Readings were taken along three row sections for each cultivar × treatment combination by moving the horizontally held light scanner below the main supporting wire at intervals of 10 cm and a total length of 2.1 m. Each row section encompassed two adjacent vines. Therefore, the total number of measuring points per row section was 22 × 64 = 1408. The light scanner was shifted across the row line at each measurement to include the entire canopy shadow projection. Incident PFD was retrieved from the same sensor used to monitor incoming light for the whole-canopy readings described hereafter. The TCLI per unit of time (μmol photons s⁻¹) was calculated as the summation of the difference between the reference (external) incoming light and the value for each pixel of the under-canopy shadow area (μmol photons m⁻² s⁻¹) multiplied by the grid area of each pixel (m²). The grid area per pixel was, in turn, calculated by multiplying step length (100 mm) by the space between sensors along the scanner (35 mm).

2.3. Single-Leaf Function and Cluster Temperature

To assess any possible interaction between cultivars, vineyard cover, and row side on leaf function, gas exchange variables [leaf assimilation rate (A), leaf transpiration rate (E), and stomatal conductance (g_s)] were measured on DOY 225 (August 13, post-veraison, TSS ∼ 10°Brix) at three different timings during the day: 12:00 to 13:30; 14:15 to 15:15; and 16:10 to 17:05. Readings were taken on three unchambered vines for each cultivar × treatment combination and, within each vine, six leaves were measured, three per row side. To account for leaf age variation along the shoot, for each row side, a basal, median, and apical leaf was measured using an LCi T Pro (ADC Bioscientific Ltd., Hoddesdon, Herts, UK). Leaves were measured at ambient CO₂ and relative humidity without altering their natural position. The same leaves sampled for gas exchange measurements were then subjected to measurements of the photochemical quantum yield of photosystem II (φPSII) taken using a leaf porometer/fluorometer (model LI-600, LI-COR, Lincoln, NE, USA). Finally, the same leaves were destructively processed for midday leaf water potential (Ψ_MD) readings. They were swiftly enveloped with a transparent bag to reduce transpiration. They were cut at the petiole with a razor blade before quickly inserted into the pressure chamber with a well-protruding petiole stub. Pressurization was stopped when the first sign of xylem sap was seen at the cut surface. Cluster temperature was measured at the three specified times of DOY 225 on three clusters per canopy side of each tagged vine, with an infrared thermometer (model Raynger ST, Raytek, Santa Cruz, CA, USA).

2.4. Whole-Canopy Gas Exchange

Whole-canopy net CO₂ exchange rate (NCER) and transpiration (T) were measured using the modified enclosure chamber method described in [25]. In brief, the system features centrifugal blowers (Vorticent C25/2 M Vortice, Milan, Italy) collecting air from a standard 1000 L plastic tank and delivering a maximum airflow of 950 m³ h⁻¹ to fill the chambers; 12 transparent polyethylene chambers (3 per each cultivar × treatment combination) were crafted around the canopy allowing 88% light transmission and 6% diffuse light enrichment and no alteration of the light spectrum (Figure S1); a CIRAS 3-DC CO₂/H₂O differential gas analyzer (PP-Systems, Amesbury, MA, USA); and a CR1000 data logger wired to an AM16/32B Multiplexer (Campbell Scientific, Shepshed, England). Every complete reading cycle lasted 16 min since the 12 solenoid valves commuted air flow from one chamber to another every 80 s. To keep the airflow fed to the chambers the most constant, 12 circular regulators with adjustable constant flow RAD2 100 BP 15–50 m³ h⁻¹ (France Air Italia, Monza Brianza, Italy) were inserted into the air tube before the air inlet of each chamber. The airflow was set at 13.8 L s⁻¹ for all chambers and maintained unaltered throughout the measurement period. Temperature was measured by shielded 1- or 0.2-mm-diameter perfluoroalkoxy Teflon-insulated Type-T thermocouples (Omega Engineering Stamford, CT, USA) at the air inlet and each chamber’s outlet. In addition, two BF-5 sunshine sensors (Delta-T Devices, Cambridge, England) were placed on top of vertical support, protruding about 50 cm above the chambers’ outlets in the OF and the covered areas for continuous measurement of direct and diffuse radiation.

Each vine was enclosed in a chamber on DOY 221 (August 9) and operated 24/7 until DOY 272 (September 29) with minimal interruptions due to either damage caused by heavy thunderstorms with strong winds (i.e., DOY 256) or minor maintenance interventions. Thus, whole-canopy measurements lasted for 51 days. Whole-canopy NCER (μmol s⁻¹) and T (g hours⁻¹) from airflow and CO₂ and H₂O differentials were calculated according to [26]. Daily gas exchange means were calculated by averaging data from dawn to dusk.

At harvest, all main and lateral nodes were counted on each per vine, and three shoots per vine were sampled and brought to the laboratory. The LA on each node of the main and lateral shoots was measured with a LA meter (LI-3000 A, LI-COR Biosciences, Lincoln, NE, USA). Afterward, the average LA was calculated for main and lateral shoots. The total LA per vine was calculated by multiplying the node number by the average leaf blade area. Since the gas exchange readings started at a time (August 9) when canopies had by far reached their final size, all gas exchange data were given on a per LA basis for normalization against absolute LA development.

2.5. Ripening Curves, Yield, and Grape Composition at Harvest

The progress of berry ripening as total soluble solids (TSS), pH, and titratable acidity (TA) was assessed in both cultivars starting at veraison (about 5 Brix in CS and about 8 Brix in MC) over 8 and 6 dates, respectively, until harvest was performed in the OF plots. Harvest dates were decided based on the stability of TSS and TA values in the OF vines; conversely, AV harvest was postponed 17 and 12 days in CS and MC, respectively. At each sampling date other than the harvest date, three 100-berry samples per cultivar × treatment combinations were taken, with equal partitioning over the two opposite row sides. Once brought to the laboratory, each sample was immediately weighed and manually pressed at room temperature, and the resulting must be used to determine TSS as Brix, pH, and TA as g L⁻¹. TSS was assessed using a temperature-compensated desk refractometer. At the same time, pH and TA were measured by titration with 0.1 N NaOH to a pH 8.2 endpoint and expressed as g L⁻¹ of tartaric acid equivalents.

At each harvest date, cluster number and total grape weight were recorded for 15 vines representative of each cultivar × treatment combinations, including the three chambered vines for each treatment in each block. Concurrently, three representative clusters per vine—usually located on shoots bore on a basal, median, and apical spur along the cordon—were taken to the laboratory for further processing. On each cluster, the severity of cluster rot, mainly as Botrytis cinerea, berry withering, and berry necrosis, was visually estimated as the percentage of affected berries over total.

From each of the 3 clusters, a 20-berry subsample was taken by carefully cutting each berry at the pedicel with sharp scissors and immediately weighed. In CS, the extraction of ATs was conducted as described by [27]; the 20 berries were peeled, and the skins were soaked in 100 mL of methanol for 24 h in a dark room at 20°C. The analyses were conducted with a spectrophotometer (Libra S80, Biochrome, Cambridge, UK), and an external calibration curve with malvidin-3-glucoside chloride as the standard (Sigma-Aldrich, St. Louis, MO, USA) was used to quantify total ATs at 520 nm. The remainder of each cluster sample was crushed, and the resulting musts were analyzed for technological maturity parameters according to the aforementioned methodology.

An aliquot of the grape juice was diluted four times and filtered through a 0.22-μm polypropylene syringe filter before the analysis to assess tartaric and malic acid concentrations. The HPLC separation was performed using an Agilent 1260 Infinity Quaternary LC system (Agilent Technology, Santa Clara, CA, USA) with a diode array detector (DAD). The chromatographic separation was performed isocratically at 0.8 mL min⁻¹ with an Allure organic acid column, 300 × 4.6 mm and 5 μm (Restek, Bellefonte, PA, USA) using 0.005 N sulfuric acid as mobile phase. The analysis was conducted at 30 ± 0.1°C, and the injection volume was 15 μL. The organic acids were detected at a wavelength of 210 nm, and the quantification was performed using external calibration with authentic standards. The analysis was performed in triplicate.

2.6. White and Red Winemaking

At respective harvest dates, three 30-kg batches of MC grapes per treatment (MC–OF and MC–AV) were manually harvested, and each sample was destemmed, crushed, and pressed with a hydraulic press to obtain approximately 25 L of juice for each batch. The juice was transferred into 30-L stainless steel vats, added with 30 mg L⁻¹ of potassium metabisulfite, and then inoculated with Saccharomyces cerevisiae at 30 g hL⁻¹ (Laffort, Milano, Italy). The fermentation was carried out at 16 ± 2°C and monitored daily by measuring the residual sugars until the end of alcoholic fermentation. At the end of the fermentation, the wines were racked, added to 40 mg L⁻¹ of potassium metabisulfite, and bottled in 750-mL glass crown-capped bottles. Winemaking of CS followed the same protocol until the addition of 30 g hL⁻¹ of Saccharomyces cerevisiae (Laffort, Milano, Italy). The fermentations were performed at 23 ± 2°C, and the pomace was manually punched down twice daily. At the end of the alcoholic fermentation, the red wines were racked off, added with 40 mg L⁻¹ of potassium metabisulfite, and bottled in 750-mL glass crown-capped bottles. The white and red wines were stored at 8°C for 2 months before analyses.

2.7. Wine Analysis

TA, pH, volatile acidity (VA), sulfur dioxide (free SO₂ and total SO₂), total phenol index (TPI), total phenols (TFs), and cinnamic acids (CAs) content were determined according to the official OIV methods [28]. Glucose and fructose were determined through an automatic multiparametric analyzer (Hyperlab, Steroglass Srl, Italy). The organic acid profile (tartaric, malic, citric) was determined as already reported by [29]. The density and ethanol content was measured with an Alcolyzer (DMA 4500, Anton Paar® GmbH, Graz, Austria). Proanthocyanidins (PROs), total ATs, flavans reactive with vanillin (FrV), and total flavonoids (TFs) were determined as reported by [30]. The colorimetric features of wines were assessed using the CIELab color space analysis, a uniform three-dimensional space defined by the colorimetric coordinates L^∗, a^∗, and b^∗. Within the CIELab space, a psychometric index of lightness, L^∗ (ranging from 0, black, to 100, white), and two-color coordinates, a^∗ (a^∗ > 0 red, a^∗ < 0 green) and b^∗ (b^∗ > 0 yellow, b^∗ < 0 blue), are defined. From these coordinates, other color parameters are defined: The hue angle (H^∗) is the qualitative attribute of color, and the chroma (C^∗) is the quantitative attribute of colorfulness: This saturation metric (S^∗) is simply the chroma divided by lightness (C^∗/L^∗) [31]. All measurements were carried out in triplicate by a V-730 UV–Vis spectrophotometer (Jasco Europe Srl, Cremella, Italy).

2.8. Volatile Composition of MC and CS Wines

Volatile compounds, including free and glycosylated terpenes, were analyzed following the method proposed by [32] with some modifications. A 100 mL of degassed wines (1:1 water: wines) added with 60 μL 2-octanol (400 mg L⁻¹) were loaded on a 1-g C18 Sep Pack cartridge (Isolute ENV+) previously activated with 5 mL methanol followed by 10 mL of deionized water. After sample deposition, cartridges were washed with 20 mL of deionized water to remove acids and sugars. Free volatiles were eluted with 10 mL dichloromethane and recovered in Sovirel vials containing anhydrous sodium sulfate before being concentrated to 200 μL under a nitrogen stream. This fraction was used later to determine free-volatile compounds. To collect glycosylated terpenes, 20 mL methanol was loaded into the previous cartridges, and the eluted fraction was dried in a rotary evaporator under a vacuum to remove methanol. The residue was dissolved in 6 mL of phosphate–citrate buffer (0.1 M Na₂HPO₄ and 50 mM citric acid; pH 5) and hydrolyzed by adding 2.6 mL of a 250 mg L⁻¹ β-glycosidase enzyme solution (Cytolase, 3000 U/g). Samples were incubated at 40°C for 16 h. To extract hydrolyzed compounds, 20 μL of 2-octanol (400 mg L⁻¹ in absolute ethanol) was added to the hydrolyzates as the internal standard and centrifuged for 15 min at 1500 g. Samples were eluted through cartridges (0.5 g C18 Sep Pack (Isolute ENV+)) previously activated with 5 mL methanol and 10 mL water. Volatiles were recovered with 5 mL dichloromethane, dehydrated with anhydrous sodium sulfate, and concentrated to 200 μL using a gentle nitrogen stream. The analysis was performed with a GC × GC system (Shimadzu Nexis GC-2030) coupled with a TQ8040 NX mass detector. The chromatographic separation was performed on two columns, with an apolar first dimension (1D) column SLB-5 ms fused silica capillary column (Supelco, 20 m × 0.18 mm × 0.18 μm film thickness) coupled to a (2D) SupelcoWAX (Supelco, 5 m × 0.32 mm dc, and df 0.25 μm film thickness). The injector, transfer line, and ion source temperature were 260°C. Oven temperature program conditions were as follows: initial temperature of 40°C for 2 min, programmed at 6°C min−1°C to 220°C. The modulation period was set to 4 s with a hot pulse time of 0.4 s. The modulator was offset by 20°C. Helium (Sapio Spa, Monza, Italy) was used as carrier gas at a constant flow of 15.5 mL min⁻¹ (0.60 mL min⁻¹ in the column). The MS parameters included electron ionization at 70 V with the ion source temperature at 260°C, mass range of 41–360 m/z, and acquisition rate of 8 spectra s⁻¹.

Compounds were identified using the following criteria: (i) by comparing their mass spectra and retention time with those of authentic standards; (ii) the compounds lacking standards were putatively identified by matching their respective mass spectra with those present in commercial libraries (NIST 2020/Wiley 12); (iii) matching the linear retention index (LRI) obtained in our conditions with the published LRI on comparable D1 columns. The volatiles were quantified by area comparison against internal standards (2-octanol) in total ion current (TIC). Bidimensional chromatograms were processed using Chrome Square 2.1 software (Chromaleont Srl, Messina, Italy). The detected peaks of interest were integrated using the same software with a custom integration method with the following conditions: similarity > 80% and ΔLRI < 30.

2.9. Chemicals and Reagents

Methanol, ethanol (LiChrosolv®), dichloromethane (Suprasolv® grade), sulfuric acid, hydrochloric acid, gallic acid, catechin, vanillin, caffeic acid, Folin–Ciocalteu reagent, sodium hydroxide, tartaric acid, malic acid, citric acid, 2-octanol, sodium sulfate anhydrous, polyvinylpolypyrrolidone) (PVPP), iodine, starch, iron sulfate heptahydrate, sulfur dioxide, and phosphoric acid were purchased from Sigma-Aldrich (St. Louis, MO, USA), and the chemicals were all at least of analytical grade. HPLC-grade water was obtained by a Milli-Q system (Millipore Filter Corp., Bedford, MA, USA).

2.10. Statistical Analysis

Within each cultivar, data were generally subjected to a one-way ANOVA within the XLSTAT package (Version 2024.2.2, Addinsoft, New York, NY, USA), and in case of the significance of the Fisher test, OF versus AV means were separated with a t-test at p < 0.05. As per single leaf and water status readings, thanks to the availability of subreplicates per each vine, the analysis was elevated to a factorial design where variety (V), vineyard treatment (T), and exposed row side (R) were the factors, in turn, generating three first-order interactions. The V × T, V × R, and T × R interactions were partitioned only when significant at p = 0.05. For the whole-canopy gas exchange data taken at different time steps on the same individuals during the season, the repeated-measures ANOVA procedure embedded in the XLSTAT software package was used. Pair comparisons within dates were conducted using the least squared mean method at p = 0.05. Regression analyses and R² calculations were also performed using the same XLSTAT package.

3.1. Weather Trends and Vineyard Microclimate

The measuring season is characterized by quite hot and dry July and August months, with T_max often peaking above 35°C (Figure S2). Conversely, September was unseasonably cool and wet: taking as a reference veraison dates of the two cultivars (July 19 and 31 for MC and CS, respectively) cumulated precipitation until respective harvests dates in OF and AV were 26.8 and 58.5 mm for MC, respectively, and 100.7 and 152.3 mm for CS (Figure S2).

Seasonal trend of mean diurnal direct radiation (PAR as μmol m⁻²s⁻¹) measured by the BF-5 hemispherical sensor under panels and in OFs covers the total length of whole-canopy gas exchange readings (DOY 221-272) (Figure 1(a)). When pooled over the 51 measuring days, direct PAR under OF conditions was 683 μmol m⁻²s⁻¹ versus 361 μmol m⁻²s⁻¹ registered under the AV cover, corresponding to a significant light depletion of 47.2% (Table 1). Daily VPD was higher in July and August versus September and peaked at about 2.5 kPa. When the same patterns were evaluated for the diffuse light component (Figure 2(a)) under panels, this amount was reduced by 49.0% compared to the OF conditions (Table 1).

Diurnal variation (from dawn to dusk) of direct radiation averaged over 51 measuring days shows that AV values are always lower than OF values, albeit with substantial intraday variation (Figure 2(a)). Under AV, a first shading “shoulder” is recorded in the morning from about 8:00 to 9:00. In contrast, vines under panel covers become heavily shaded just before solar noon (about 2 p.m. according to the specific site location and row orientation) until 4 p.m., when shading is alleviated before light starts to decline in both treatments with approaching sunset. The diurnal pattern of diffuse light under AV averaged over the same time frame is smoother, showing the lowest diffuse-to-direct light fractions (37%–41%) when the sun is around the zenith point (Figure 2(b)). When the diurnal patterns of air temperature measured at the chambers’ outlets were plotted, it was apparent that, in both cultivars, maximum cooling occurred around 10 a.m. with about 1.6°C–1.7°C less under AV concerning OF (not shown).

The presence of the AV panels profoundly changed patterns of both incoming and intercepted light (Table 1). Generally, fractions of intercepted light to incoming light were much higher than under OF, peaking at 75.3% for readings taken in the morning under the MC–AV. However, these ratios also reflect significant changes in incoming radiation under panels (about 72.6% less vs. OF for data averaged over cultivars and daytime).

3.2. Whole-Canopy Gas Exchange

Seasonal net carbon exchange rates (NCERs) per unit of LA are shown in Figure S3 for CS (A) and MC (B). Repeated ANOVA has shown neither significant main treatment effects nor time × treatment interactions in both cases. In CS, NCER/LA averaged over the 51 measuring days yielded 3.56 µmolm⁻²s⁻¹ in AV and 3.91 μmol m⁻²s⁻¹ in OF for a 9.2% seasonal NCER reduction under panels. In MC, the same comparison gave 3.71 and 3.96 µmolm⁻²s⁻¹ for AV and OF, respectively, corresponding to a 6.3% NCER reduction under panels.

In both cultivars, seasonal T/LA varied between 0 and about 100 g m⁻²h⁻¹, and the system was able to catch the wide fluctuations occurring on cool and cloudy days versus warm and sunny days (Figure S4). Once again, repeated ANOVA did not result in any significant main treatment and time × treatment effects. Once averaged over the whole measuring season, CS T/LA was 44.6 g m⁻²h⁻¹ in AV and 48.4 g m⁻²s⁻¹ in OF, reducing 7.9% under cover (Figure S4A). In MC, average T/LA rates were 50.4 g m⁻²h⁻¹ in AV and 54.3 g m⁻²s⁻¹ in OF, for a 7.2% reduction under panels (Figure S4B).

Relative variation of NCER/LA and T/LA over the season originated in CS, an almost identical canopy WUE (5.30 µmolCO₂m⁻²s⁻¹/mmolH₂Om⁻²s⁻¹ in AV vs. 5.33 µmolCO₂m⁻²s⁻¹/mmolH₂Om⁻²s⁻¹ in OF) (Figure S5A). Similar behavior was apparent from the MC, where canopy WUE was set at 4.85 µmolCO₂m⁻²s⁻¹/mmolH₂Om⁻²s⁻¹ in AV versus 4.74 µmolCO₂m⁻²s⁻¹/mmolH₂Om⁻²s⁻¹ in OF (Figure S5B). Unsurprisingly, repeated ANOVA carried out on both datasets did not highlight significant main treatment and time × treatment interaction.

When evaluated on a diurnal basis with data averaging over the 51 measuring days, NCER/LA showed trends reported in Figure 3 for CS (A) and MC (B). In CS, a significant time × treatment interaction occurred (p = 0.001), indicating that, in AV, NCER/LA was significantly lower than that recorded in OF at 8 timings during the day which was concentrated in the morning until about 10 a.m. and in the early afternoon (from 2 to 4 p.m.). Although in MC similar trends were appreciated (Figure 3(b)), the same never reached statistical significance.

Diurnal trends of T/LA in the two cultivars mimicked those already described for NCER/LA (Figure 4). However, in CS significant differences between AV and OF were limited to 5 timings early in the morning, whereas no statistical differences were detected over the remainder of the day. This concurs with a significant time × treatment interaction (F = 2.307, p = 0.001) highlighted by the repeated ANOVA. In Malvasia, similarly to the NCER/LA rates, no differences between T/LA measured on AV and OF canopies were found at any time of the day (Figure 4(b)).

Calculated diurnal canopy WUE showed, for CS (Figure 5(a)), a significant time × treatment interaction, which highlighted, around noon, two timepoints when AV had lower WUE while the opposite occurred on four timepoints over the late afternoon. The same effects did not reach significance in the diurnal canopy WUE calculated for MC (Figure 5(b)).

When light response curves were plotted for the two varieties using diurnal NCER/LA pooled over all measuring days and direct radiation as PAR, both datasets closely fitted to a multiparameter sigmoid model with R² values ranging from 0.88 to 0.96 (Figure 6). Regardless of cultivars, it was apparent that vines growing under AV had a lower light saturation point (about 600 μmol m⁻²s⁻¹) than that recorded in OF and approaching 1000 μmol m⁻²s⁻¹; moreover, at intermediate PAR levels (about 200–500 μmol m⁻²s⁻¹) in both cultivars, vines grown under AV had higher NCER/LA than the OF vines.

3.3. Single-Leaf Gas Exchange, Function, and Water Status

Single-leaf gas exchange and water status data recorded late in the morning of DOY 225 (August 13) are reported in Table 2 as the main effects (cultivar, type of cover, row side) and relative one-order interactions. While the slightly better leaf function recorded in CS seems to be related to a higher amount of incoming radiation during the time measurements were taken, it is relevant that the three parameters affected by the presence of AV were cluster temperature, ϕPSII, and Ψ_MD which all pointed at less stressful conditions versus the OF plots (Table 2). Due to the well-exposed condition of the east-facing row side at the time of measurements (incoming PAR = 1240 μmol m⁻²s⁻¹), leaf gas exchange was enhanced for most of the measured parameters. The first-order interactions were mostly significant for T_CLUSTER and ϕPSII (Table 3). As expected, panel cover significantly reduced the difference in temperature between E- and W-exposed clusters versus the OF conditions; in terms of V × R interaction, CS clusters showed a lesser T gradient when passing from E to W exposure than MC (Table 3). Panel presence reduced differences in Ψ_MD between the two varieties, which, instead, amplified under OF conditions with CS hitting the lowest value of −1.30 MPa.

First-order interactions found for (φPSII) showed that both cultivars reached suboptimal φPSII values for leaves exposed to the E side, whereas φPSII did not differ between cultivars under OF conditions. For data pooled over cultivars, the OF-E combination reached the highest limitation (0.238). In the morning, leaves inserted on the W side of the canopy had much lower WUE_inst than those placed on the E side (Table 3).

The same survey repeated at about solar noon (Table 4) produced a quite different outcome. Due to the very limited incoming radiation, between-variety differences were minor overall. Conversely, the main effects related to the presence or absence of AV cover originated a quite peculiar result. Although AV received only 152 μmol m⁻²s⁻¹, which caused lower T_CLUSTER and A (Table 4), leaf water status was maintained under AV in terms of higher E, g_s, and Ψ_MD. Due to such relative changes, WUE_inst and WUE_i dropped under panel cover. The main effect of the row side showed minimal impact on the performed readings, with the sole exception of the west-exposed row side being 1.2°C warmer than the opposite row side. The 2 T × R and V × R significant interactions found again for T_CLUSTER were of low magnitude, whereas the V × T interaction found for Ψ_MD very closely mirrored trends observed in the morning (Table 3).

Data recorded in the afternoon confirmed minor differences between varieties due to limited incoming radiation (Table 5). In terms of treatment’s main effect, at almost the same incident PAR, AV behaved better than the OF vines as it showed higher g_s and A, coupled with a less negative Ψ_MD. As per row-side effects, the east exposed was strongly energy limited (only 114 vs. 1454 μmol m⁻²s⁻¹ hitting the west-exposed side), leaf function as PSII was optimal, and WUE was also low. Finally, T_CLUSTER was confirmed to be quite responsive in terms of interactions (Table 3): Once again, the presence of panels annulled T_CLUSTER differences between E- and W-exposed rows, whereas, under OF, a 2.5°C heating for the W side was recorded. To mimic what was seen for both morning and noon measurement windows, Ψ_MD did not vary between cultivars under AV cover. In contrast, CS was confirmed to reach a more negative Ψ_MD than MC under OF conditions.

3.4. Vegetative Growth, Yield Components, and Grape Ripening

There were no significant changes in CS vines due to panel cover on the total LA and all main yield components (Table 6). Consequently, the LA/Y ratio across treatments was almost the same (about 1.6 m²/kg). However, although the fresh berry weight did not differ, berries grown under AV had a lower skin mass per berry and, consequently, a lower relative skin percentage.

Repeated-measures ANOVA was performed on the TSS and TA data gathered on different dates post-veraison. In CS, a highly significant date × treatment interaction for both parameters originated (Figure 7(a)). Ripening curves showed that soon after panel setup, AV started to significantly delay sugar accumulation and acidity degradation. When OF vines were harvested on September 24, their TSS was still higher than the value found in AV. Extending the harvest date of AV by 17 days did not allow full recovery, as TSS was set at 20.9 Brix versus 22.2 scored for the OF treatment (p = 0.005). TA decline was also slowed down in AV and, at respective harvest dates, CS vines grown under panel cover retained higher total acidity and malic concentration than OF and, instead, lower tartaric acid concentration. Total ATs were severely curtailed in AV regardless of the unit of expression (Table 6). Although berry dehydration and necrosis were minimal in the OF treatment, panel cover allowed for almost the elimination of this low percentage. Due to the extremely rainy period elapsing between the two harvest dates in CS (Figure S2), cluster rot incidence reached a remarkable 70% compared to the OF treatment.

In MC, total LA and yield per vine were very homogeneous between the two treatments, originating, in turn, a very close LA/Y ratio (about 0.9 m²/kg) (Table 6). To partially differ from the results shown in Cabernet S, AV promoted a larger berry size and cluster weight, which was also higher, although at the border of significance (p = 0.06). Repeated ANOVA performed on the TSS and TA data obtained at different dates post-veraison originated a highly significant date × treatment interaction only for TSS (Figure 7(b)), showing a delayed sugar accumulation in AV since DOY 221 that was overall maintained for the remainder of the season. However, at the harvest of OF, this gap was almost filled (17.8°Brix in AV vs. 18.5°Brix in OF, ns), and at the respective harvest dates (Table 6), TSS was very close to full compensation in AV. Nevertheless, AV retained higher pH and malic acid concentration at similar TSS. In Malvasia, panel covers proved effective at significantly decreasing the incidence of berry withering and necrosis. Albeit significant, the increase in cluster rot was inconsistent (0.8% at AV harvest).

3.5. Wine Properties

The chemical–physical compositions of MC (MC–OF vs. MC–AV) and CS (CS–OF vs. CS–AV) wines are reported in Tables 7 and 8. All alcoholic fermentations were regularly completed within 7–8 days (residual sugars < 0.2 g L⁻¹). VA and total sulfur dioxide concentrations did not show significant differences between treatments and ranged from 0.27 to 0.36 g L⁻¹ and from 64.0 to 69.0 g L⁻¹, respectively, falling within the legal limit (EU Regulation No. 606/2009). The OF wines showed the lowest pH values compared to the AV wines, while no significant difference in TA, density, and ethanol content was found among the treatments.

Table 7. Wine attributes obtained from Cabernet Sauvignon grapes.

	Unit	CS–OF	CS–AV
Wine attributes
pH	—	3.18	3.21
Titratable acidity1	g L−1	6.49	6.80
Ethanol	% (v/v)	10.89	10.01
Density	g L−1	0.99	0.99
Free SO2	mg L−1	16.3	14.0
Total SO2	mg L−1	64.3	65.0
Volatile acidity2	g L−1	0.36	0.35
Organic Acids
Tartaric acid	g L−1	2.94	2.36
Malic acid	g L−1	2.53	3.09
Citric acid	g L−1	1.46	0.87
Chromatic features
L∗	—	37.07	46.42
a∗	—	72.16	71.52
b∗	—	50.7	34.9
H∗	—	35.1	26.1
C∗	—	88.2	79.7
S∗	—	2.39	1.72
DE	—	—	2.80
Color8	—
Phenolics
Total phenol index	—	26.54	19.52
Total phenols3	mg L−1	1495.50	1259.33
Cinnamic acids4	mg L−1	396.9	322.71
Proanthocyanidins5	mg L−1	1599.0	1076.6
Total anthocyanins6	mg L−1	152.64	76.73
Flavans reactive to vanillin7	mg L−1	607.40	420.85
Total flavonoids7	mg L−1	1065.29	817.96

• Note: Within each row, values in bold are different at p ≤ 0.05 according to t-test, n = 3. CS–OF: Cabernet Sauvignon vines under open-field (OF); CS–AV: Cabernet Sauvignon vines under agrivoltaics (AV) panels.
• ¹Titratable acidity: expressed as g L⁻¹ of tartaric acid equivalent.
• ²Volatile Acidity: expressed as g L⁻¹ of acetic acid equivalent.
• ³Total phenols: expressed as mg L⁻¹ of gallic acid equivalent.
• ⁴Cinnamic acid: expressed as mg L⁻¹ of caffeic acid equivalent.
• ⁵Procyanidins: expressed as mg L⁻¹ of cyanidins acid equivalent.
• ⁶Total anthocyanins: expressed as mg L⁻¹ of malvidin equivalent.
• ⁷Flavans reactive to vanillin: expressed as mg L⁻¹ of catechin equivalent.
• ⁸Color of wine reproduced on the basis of CIELab coordinates by using the colorizer software (https://colorizer.org. accessed on 28-01-2025).

Table 8. Wine attributes obtained from Malvasia di Candia aromatica grapes.

	Unit	MC–OF	MC–AV
Wine attributes
pH	—	3.12	3.20
Titratable acidity1	g L−1	6.70	6.41
Ethanol	% (v/v)	10.42	10.55
Density	g L−1	0.99	0.99
Free SO2	mg L−1	18.67	16.33
Total SO2	mg L−1	68.00	68.67
Volatile acidity2	g L−1	0.28	0.27
Organic acids
Tartaric acid	g L−1	2.05	1.81
Malic acid	g L−1	1.33	1.42
Citric acid	g L−1	0.58	0.66
Chromatic features	—	—	—
L∗	—	100.21	102.41
a∗	—	0.18	−0.03
b∗	—	7.42	6.55
H∗	—	88.76	89.57
C∗	—	7.42	6.55
S∗	—	0.07	0.06
DE	—	—	2.00
Color8	—
Phenolics	—	—	—
Total phenol index (TPI)	—	8.1	7.31
Total phenols3	mg L−1	233.13	238.5
Cinnamic acids4	mg L−1	89.56	83.12
Proanthocyanidins5	mg L−1	29.86	42.15
Total anthocyanins6	mg L−1	nd9	nd
Flavans reactive to vanillin7	mg L−1	11.1	13.94
Total flavonoids7	mg L−1	nd	nd

• Note: Within each row, values in bold are different at p ≤ 0.05 according to t-test, n = 3. MC–OF: Malvasia di Candia aromatica vines under open-field (OF) conditions; MA-AV: Malvasia di Candia aromatica vines under agrivoltaics (AV) panels.
• ¹Titratable acidity: expressed as g L⁻¹ of tartaric acid equivalent.
• ²Volatile acidity: expressed as g L⁻¹ of acetic acid equivalent.
• ³Total phenols: expressed as mg L⁻¹ of gallic acid equivalent.
• ⁴Cinnamic acid: expressed as mg L⁻¹ of caffeic acid equivalent.
• ⁵Procyanidins: expressed as mg L⁻¹ of cyanidins acid equivalent.
• ⁶Total anthocyanins: expressed as mg L⁻¹ of malvidin equivalent.
• ⁷Flavans reactive to vanillin: expressed as mg L⁻¹ of catechin equivalent.
• ⁸Color of wine reproduced on the basis of CIELab coordinates by using the colorizer software (https://colorizer.org. accessed on 28-01-2025).
• ⁹nd: not detectable.

As reported in Table 7, the final CS wine composition was maintained for OF higher tartaric and lower malic, as previously ascertained in grape juices. Regarding the wine’s color, OF wines had a higher Hue angle (H^∗), chroma (C^∗), yellow-blue color coordinate (b^∗), and color saturation (S^∗), whereas AV wine showed a higher lightness (L^∗). TFs and all classes of phenolics were significantly curtailed under AV cover. As per volatile compounds (Table 9), several esters, monoterpenes (namely, D-Limonene, citronellol, and nerolidon), and nor-isoprenoids (-damascenone) were lower in AV wines.

Final MC wines had fairly similar basic attributes and overall confirmed must analyses (Table 8). As per phenolics components, the AV wines had lower TPI and CAs and higher PROs. Volatile compounds differed between the two wines: esters and several free monoterpenes were significantly lower in AV wines (Table 10). However, the prevailing bound aromatic compounds showed that geraniol plus derivatives concentration did not differ between the two wines, and AV doubled the content in nerol versus OF wines.