Abstract
The impact of mechanical harvesting, optical berry sorting, and their possible synergistic effect on grape and wine composition was investigated. Pinot noir grapes from the Russian River Valley American Viticultural Area were harvested by hand, by a standard bow-rod mechanical harvester, or by a mechanical harvester with a Selectiv’ Process on-board. For each harvest method, half of the grapes were unsorted and half were optically sorted at the winery. The grapes, wines at bottling, and wines after three months of bottle aging were analyzed by reversed-phase high-performance liquid chromatography (RP-HPLC), ultraviolet-visible spectroscopy (UV-vis), and headspace solid-phase microextraction (HS-SPME) gas chromatography mass spectrometry (GC-MS) for color expression and phenolic and aroma profiling. The machine-harvested grapes had higher levels of β-damascenone, linalool, β-myrcene, and α-terpinene, potentially caused by glycosidic hydrolysis triggered by berry damage during harvest or from induced synthesis as a wounding response. In general, differences in wine composition attributable to harvest method were diminished or eliminated by optical sorting. The machine harvester with the Selectiv’ Process on-board led to wines with the most phenolics, although these differences may have been decreased or eliminated had the grapes been crushed before fermentation as the wines were produced by whole-berry fermentation. Descriptive sensory analysis conducted on wines three months after bottling determined that the wines made from hand-harvested fruit had significantly greater tropical fruit aroma, while wines made from optically sorted treatments had less hue saturation. With only two significant differences among the 18 aroma, taste, and mouthfeel attributes tested, it was concluded that all treatments led to wines of similar character.
With the ever-increasing cost and shortage of qualified labor and the desire to economize vineyard operations, mechanized grape harvesting has become increasingly important for wine production. Concerns associated with mechanical harvesting include (i) physical damage to fruit resulting from the rapid shaking required to separate berries from rachis; (ii) the inclusion of undesirable second crop, overripe or moldy clusters, and material other than grapes (MOG); (iii) the increased risk of oxidation, enzymatic activity, and development of microbial populations in broken and therefore vulnerable fruit during transport from vineyard to winery; and (iv) loss of juice in the vineyard. Several studies have investigated the impact of mechanical harvesting on grape and wine composition. One study evaluated wines made with machine- and hand-harvested Chardonnay grapes and found that subjects had no significant preference between the two for both young wines and wines aged 18 months (Clary et al. 1990). Another study used both red and white varieties (Petite Sirah, French Colombard, and Chenin blanc) and found that subjects had no preference between wines made from grapes harvested by the two methods (Noble et al. 1975).
An ideal harvesting mechanism would rapidly and efficiently select fruit within desired parameters while excluding MOG. Some new mechanical harvesters include an on-board picking head that eliminates pieces of rachis, leaves, and shoots; however, these mechanisms do not exclude moldy or overripe berries. Typically, sorting at the winery eliminates undesirable fruit picked by mechanical harvesters. Sorting mechanically-harvested fruit by hand, however, is tedious and requires tremendous resources, since inspection of individual berries is necessary to sort the already destemmed fruit. Optical sorters, however, are well-suited to rapidly sort destemmed grapes. Sorting is based on parameters such as berry size, color, and shape, and also eliminates foreign material. Although optical berry sorters have become more common in commercial wine production, there are relatively few studies investigating their effects on the chemical and sensory properties of wine. Wines made from optically sorted Chardonnay grapes had higher pH and more total phenols and residual sugar than unsorted controls (Falconer et al. 2006). Sensory analysis showed that wines made from sorted berries had more tropical fruit aroma and sweetness. However, these wines had no significant differences in other sensory attributes and the authors concluded that the wines were similar in character (Falconer et al. 2006). In a study using early optical berry sorting technology, Muscadine grapes were sorted into four groups using absorbance parameters (Carroll et al. 1978). The grapes were processed and chemical analyses showed that Brix and pH increased with successive sorting (ripeness) levels, while titratable acidity decreased. In the resulting wines, pH and tannin levels increased and titratable acidity decreased with successive sorting levels. A sensory panel evaluated the finished wines and found that the wines made from the first and fourth sorting groups were inferior to those made from the two middle groups, which were deemed to have optimal ripeness. Although this study employed now-outdated technology, it shows that optical berry sorting can successfully segregate grapes by ripeness with chemical and sensory impacts on resulting must and wine.
There have been few studies regarding the effect of mechanical harvesting, and even fewer regarding optical berry sorting. Furthermore, no previous studies that address possible interaction effects of these two techniques were found. Although the limited studies available indicate a relatively small impact of either mechanical harvesting or optical berry sorting, the general perception by the grape and wine industry is that mechanical harvesting negatively impacts quality. Additionally, optical berry sorters are promoted to remove the potential impact of harvest method on grape composition. The aim of this study was to determine the individual and synergistic effects of mechanical harvesting and optical berry sorting on grape and wine composition.
Materials and Methods
Harvest methods
The grapes in this study were sourced from a commercial Vitis vinifera L. cv. Pinot noir vineyard (clone Dijon 667 grafted on 1103 Paulsen rootstock) located in the Russian River Valley American Viticultural Area, California. The vines were spaced 2.4 m × 1.5 m, cane-pruned, and bilateral cordon-trained on a vertical shoot-positioned (VSP) trellising system. The grapes were harvested just before sunrise on 17 Sept 2013 using three methods. The grapes were in good condition with no visible rot or mold. There was low disease pressure during the 2013 season due to the prolonged drought in California. One metric tonne (1000 kg) of grapes was picked by hand, one tonne was mechanically harvested with a Pellenc overrow tractor 8590 with a Selectiv’ Process on-board picking head (Pellenc America; henceforth referred to as “Selectiv’”), and one tonne was mechanically harvested with the same Pellenc harvester with the on-board picking head disengaged, which thus operated like a standard bow-rod machine harvester (“machine”). The Selectiv’ Process on-board picking head does on-board sorting of harvested grapes, removing mostly MOG with all grapes accepted. The grapes were harvested from four adjacent vineyard rows. The harvesting methods were rotated periodically during the process to minimize row-to-row variation. The grapes were transported in half-ton bins to the UC Davis Teaching and Research Winery.
Optical berry sorting
For each harvesting treatment, half of the grapes received no sorting at the winery and half were sorted with a 2011 Delta Vistalys R1 optical sorter (Bucher-Vaslin) set to a stringency level of four out of five, resulting in six total treatments. The optical sorter was prepared by sending a sample of hand-selected “ideal” berries through the machine while running the training mode to establish rejection parameters. To minimize processing variability among treatments, grapes from all lots were sent through a Delta E2 destemmer (Bucher-Vaslin) and the optical sorter, with the optical sorting mechanism turned off for the treatments that were not sorted. The rejection rate was 9 ± 1% of the fruit based on a weight per weight basis.
Berry sampling
Three 500-berry grape samples were selected randomly from each treatment directly following de-stemming and movement through the optical sorter (whether in use or not), and stored at −20°C until analysis within six months.
Winemaking
Fermentations were carried out in triplicate using 200-L stainless steel fermentation vessels. Fermentors were filled incrementally to ensure homogeneity of fruit in each ferment replicate (122.1 ± 6.6 L). Fifty mg/L sulfur dioxide was added in the form of 15% potassium metabisulfite solution directly after filling and mixed by stirring. The morning after receiving the grapes, 0.24 g/L Fermaid K (Lallemand) was added. The diammonium phosphate (DAP) needed to achieve total yeast assimilable nitrogen levels of 300 mg/L was added in two installments: before inoculation and after one-third sugar depletion. The musts were inoculated 36 hr after arrival at the winery with Saccharomyces cerevisiae strain Lalvin D254 (Lallemand) according to the manufacturer’s rehydration procedure. Fermentation temperatures were controlled at 22 ± 1°C. Cap management was addressed by pumping one tank volume of wine from the bottom of the tank over the fermentations through an irrigator. The first pump-over took place 12 hr after inoculation and continued three times per day (one aerative) until the wines reached five Brix, then twice per day (one aerative) until the wines were dry. Wines were pressed using a basket press and returned to the research fermentors to settle. Five days after pressing, the wines were racked off lees into 49.2-L stainless steel drums and moved to a 20°C storage room. Wines were inoculated for malolactic fermentation (MLF) with Viniflora Oenococcus oeni (Chr. Hansen A/S) according to the manufacturer’s protocol. MLF were monitored weekly by following the decrease in malic acid concentration using a Gallery Automated Photometric Analyzer (Thermo Fisher Scientific). After completion of MLF, 50 mg/L sulfur dioxide was added and the tanks were transferred to a 9°C storage room. All treatments fermented similarly and completed MLF within the same week. The wine chemical compositions were determined at time of bottling for all treatments. The % ethanol (v/v) was measured with an alcolyzer (Anton Parr). The pH was measured using an Orion 5-star pH meter (Thermo Fisher Scientific). The total acidity (TA) was measured automatically with the Mettler-Toledo DL50 titrator (Mettler-Toledo, Inc.). The measurements of acetate, malate, and residual sugar (RS) were made using the Gallery automated analyzer (Thermo Fisher Scientific).
Chemicals and reagents
All chemicals and reagents used were purchased from Sigma-Aldrich with these exceptions: malvidin-3-O-glucoside chloride (Indofine Chemical Company, Inc.), ethyl 2-methylbutyrate (SAFISIS), linalool (Alfa Aesar), acetic acid (EMD, Merck), hexanoic acid (Acros Organics), nerol oxide (Penta Manufacturing), and 100% ethanol (Gold Shield Chemical). All standards were of the highest purity available.
Phenolic extraction and analyses
Twenty grapes were selected at random from each frozen berry sample replicate and weighed. The berries were homogenized for 2.5 min with 1.0 mL 50% v/v ethanol solution with 0.1% v/v hydrochloric acid and 0.1% w/v ascorbic acid for every 0.1 g of berry tissue, using an Ultra-Turrax T 18 basic disperser (IKA Works, Inc.) to extract anthocyanins and other monomeric phenols. The mixture was placed in 50 mL centrifuge tubes and refrigerated for 16 hr with occasional agitation, then centrifuged (Eppendorf Centrifuge 5810R) at 3200 rpm for 15 min. The supernatant from each sample was stored at −20°C. The pellet was suspended in acidified acetone (70:30 acetone:water with 0.1% ascorbic acid (w/v) using the same volume as for ethanol extraction to extract larger polymers. The acetone extractions were centrifuged as before after 16 hr at 4°C with occasional mixing. The ethanol and acetone supernatants were combined and reduced under vacuum at 35°C to near dryness using a Rotavapor R110 (Buchi), and made up to 15 mL in a volumetric flask using model wine (14% v/v ethanol, 5 g/L potassium bitartrate, adjusted to pH 3.4 with 37% hydrochloric acid) and kept at −20°C until analysis.
Absorbance (A) from 200 to 900 nm of each wine sample was measured using a Hewlett-Packard 8453 UV-vis Spectrophotometer with a quartz 100 μm path-length flow-cell (Starna Cells). Prior to analysis, frozen samples were thawed at room temperature and centrifuged at 3200 rpm for five min at 4°C. Color hue was the ratio of 420 to 520 nm, while color density was defined as the sum of A at 420, 520, and 620 nm (Somers and Evans 1977, Ribéreau-Gayon et al. 2006).
The polyphenol compositions of the grape and wine samples were determined using the modified Adams-Harbertson assay (Harbertson et al. 2003). The method determined total iron-reactive phenolics and tannin expressed as catechin equivalents (CE), and anthocyanins expressed as malvidin-3-glucoside (M3G) equivalents.
Phenolics were profiled by reversed-phase high performance liquid chromatography (RP-HPLC). Chromatographic separations were performed on an Agilent 1260 Infinity HPLC equipped with a thermostatted autosampler, thermostatted column compartment, and diode array detector. An Agilent Poroshell 120 SB-C18 (4.6 × 150 mm, 2.7 μm particle) column at 35°C was used for all analyses. A gradient separation was performed using an aqueous solution of 5% formic acid (solvent A) and acetonitrile containing 10% (v/v) solvent A (solvent B) at a flow rate of 1.25 mL/min. The elution conditions were as follows: 0 min, 5% solvent B; 23 min, 27% B; 24 min, 95% B; 26 min, 95% B; and 26.5 min, 5% B. The column was allowed to equilibrate for 6.5 min prior to the next injection. Injection volume for all samples was 20 μL. A diode array detector was used to monitor phenols of interest at wavelengths of 280 nm (gallic acid, (+)-catechin, (−)-epicatechin, and tannins), 320 nm (caftaric, coutaric, and caffeic acids), 370 nm (quercetin and quercetin glycosides), and 520 nm (anthocyanins). Identification of the compounds eluting from the RP-HPLC column was based on comparison to authentic standards where available; otherwise, according to spectral absorbance and expected retention time published in the literature (Peng et al. 2002). Chromatograms were integrated using Agilent CDS ChemStation Software. Gallic acid, (+)-catechin, caffeic acid, p-coumaric acid, quercetrin, quercetin, and malvidin-3-O-glucoside were quantified using calibration curves produced with authentic standards of each compound with limits of quantification of 0.50, 1.00, 0.25, 0.25, 0.50, 1.00, and 0.50 mg/L, respectively. (−)-Epicatechin and tannins were quantified as catechin equivalents, caftaric acid was quantified as caffeic acid equivalents, coutaric acid was quantified as p-coumaric equivalents, all quercetin glycosides were quantified as quercetrin equivalents, and all anthocyanins and pigmented polymers were quantified as malvidin-3-O-glucoside equivalents.
Analysis of aroma compounds by HS-SPME-GC-MS
Volatile aroma compounds were analyzed in grape samples, in wines at bottling, and in the wines after three months aging using automated headspace solid phase microextraction (HS-SPME) gas chromatography-mass spectrometry (GC-MS). The samples were analyzed using a 6890 gas chromatograph coupled to a 5975 mass selective detector (MSD) (Agilent) equipped with an MPS2 autosampler (Gerstel). Sixty-three compounds were selected for analysis in grape and wine samples (Table 1) based on previous reports on the aromatics of Pinot noir (Moio and Etievant 1995, Fang and Qian 2005, Fang and Qian 2006) with the addition of compounds that were seen in samples, but not mentioned in the above studies. All of the compounds were verified by analyzing reference standards as previously described (Hjelmeland et al. 2013) with the following additions: ethyl cinnamate, ethyl vanillate, methyl vanillate, acetovanillone, nerol oxide, ethyl anthranilate, and methyl anthranilate. These reference chemicals had purities ≥99% except for acetovanillone, which had a purity ≥98%, and ethyl vanillate and ethyl cinnamate, which had unverified purities.
Compounds measured by HS-SPME-GC analysis in grapes, wines at bottling, and wines after three months aging, with their CAS number, retention indices (RI), and source for DB-Wax, retention time, calculated retention indices (CRI), and selected ion monitoring (SIM) qualifying ions.
The grape samples were prepared for volatile analysis by an adaptation of a procedure described previously (Canuti et al. 2009). Samples were kept on ice as much as possible during the extraction protocol. Twenty g (± 0.20) of frozen, randomly selected grapes from each sample replicate was weighed and placed into 50 mL polypropylene Falcon tubes (Corning Life Sciences) containing 2 mL buffer solution (1.0 M sodium citrate dihydrate, pH 6.0), 20 μL ascorbic acid solution (200 g/L prepared in millipore water), and 50 μL 2-octanol internal standard solution (10 mg/L prepared in 100% ethanol) to account for sample loss. Sample mixtures were homogenized with an Ultra-Turrax T 18 basic disperser (IKA Works, Inc.) and then centrifuged (Eppendorf Model 5403) at 4°C for 10 min at 5000 rpm. Eight mL aliquots of the supernatant were transferred to 20 mL amber glass headspace sampling vials containing 3 g (± 0.10) NaCl (Fisher Scientific). Fifty μL 2-undecanone solution (10 mg/L prepared in 100% ethanol) was added to the vials (beneath the sample surface to prevent volatilization) as a second internal standard. Vials were sealed with 1 mm thick magnetic screwcaps with PTFE-lined silicone septa (Supelco). Samples were analyzed in triplicate in an alternating order within nine hours of loading on the instrument and normalized to both internal standards. Wine samples were prepared in vials similarly to the grape extracts.
The HS-SPME-GCMS procedure used for the grape samples was adapted from a previously described method (Hjelmeland et al. 2013). A 1 cm polydimethylsiloxane (PDMS) (Supelco), 23 gauge SPME fiber was used for sampling. Samples were warmed at 30°C and agitated at 500 rpm for five min before exposing the fiber for 45 min at 30°C with agitation at 250 rpm. A DB-Wax ETR capillary column (30 m, 0.25 mm i.d., 0.25 μm film thickness; J&W Scientific) and SPME inlet liner (0.7 mm i.d.; Supelco) were used. The instrument was controlled by Maestro (ver. 1.2.3.1, Gerstel) and the data was analyzed using ChemStation software (ver. E.01.01.335; Agilent). Helium at a constant pressure of 6.69 psi was used as the carrier gas. The method was retention time locked to the internal standard, 2-undecanone, at constant pressure to prevent retention time drifting. During analysis, the oven was kept at 40°C for five min, then increased 3°C/min up to 180°C, and then 30°C/min up to 260°C, before holding for 7.67 min. The MSD interface was held at 260°C. The inlet temperature was held at 260°C and the SPME fiber was desorbed in splitless mode with the split vent opening at 1.2 min. The solvent delay was two min and the detector was turned off from 3.80 min to 4.30 min during ethanol elution. The fiber was held in the inlet for 10 min to prevent carryover effects. Electron ionization source was used, with a source temperature of 230°C and the quadrupole at 150°C. The samples were measured using synchronous scan and selected ion monitoring (SIM mode). The scan range was from 40 m/z to 300 m/z and compounds were detected using between two and six selected ions with a scan rate of 5.80 scans/sec. The ions used in the SIM parameters and retention times for each compound are shown (Table 1).
For the wine samples, a 1 cm polydimethylsiloxane/divinylbenzene/carboxen (PDMS/DVB/CAR; Supelco), 23 gauge SPME fiber was used for sampling. Samples were warmed at 30°C and agitated at 500 rpm for five min before exposing the fiber for 45 min at 30°C with agitation at 250 rpm. The chromatograph, autosampler, column, inlet liner, and software used and separation conditions were identical to those described for grape analysis except for the details listed below. Helium at a constant pressure of 6.55 psi was used as the carrier gas. The SPME fiber was desorbed in split mode with a 10:1 split ratio.
Descriptive sensory analysis
The sensory profiles of the Pinot noir wines were analyzed approximately three months after bottling using descriptive sensory analysis (DA) in the J. Lohr Wine Sensory Room, University of California, Davis, CA. Thirteen panelists were recruited by advertising within the university, none of whom had a reason not to consume alcohol. The study was approved by the Institutional Review Board of the university (IRB ID 571920-1) and all panelists gave informed oral consent and received no financial compensation for their participation. None of the panelists were aware of the purpose of the study or how many different samples they were evaluating. Two fermentation replicates of each treatment were randomly selected for DA, totaling 12 wines. Training for the panel consisted of five one-hr sessions over two wks. The research wines were presented blindly to the panelists, who generated a comprehensive list of attributes. The list was reduced through group discussion and consensus, and reference standards for each attribute were created for concept alignment and training. The panel agreed on 12 aroma attributes, five taste and mouthfeel descriptors, and one visual assessment of hue saturation intensity (Table 2). Wines were evaluated by rating attributes on an unstructured line scale anchored by the words “low” and “high.” The panelists were asked to use the whole scale by defining the extremes of the scale relative to the minimum and maximum intensities of attributes found in the wines.
Attributes used for descriptive sensory analysis and corresponding reference standards. Unless otherwise stated, reference standards were prepared using 30 mL of experimental wine as a starting base.
Wines were presented in triplicate in a randomized block design, with six wines per session over three wks in isolated sensory booths equipped with positive air flow and white lights. Wine samples (30 mL) were identified by random three-digit codes and presented in black tasting glasses (ISO-3591:1977) to eliminate biases introduced by possible color differences. To minimize palate fatigue, samples were separated by a forced one min break, with a five min break between wines three and four, during which panelists evaluated the hue saturation intensity in a Macbeth light box. Purified water was provided as a palate cleanser and all samples were expectorated. Sensory data were collected using FIZZ software (ver. 2.00L, Biosystemes).
Statistical analysis
All chemical and sensory data were tested for statistical significance using multivariate analysis of variance (MANOVA) for the overall main treatment effect. If significant, univariate analyses of variance (ANOVA) measuring for the effects of treatment and replicate using a pseudo-mixed model test was used for all chemical data. For the DA, an ANOVA using the effects of judge, treatment, and replicate with a pseudo-mixed model was used. Fisher’s least significant differences (LSD) were calculated to assess significance groups. For the GC-MS data of aroma compounds, treatments were compared graphically using principal component analysis (PCA) on averaged data. The chemical and descriptive sensory data were related to one another using a partial least squares regression (PLSR). Statistical significance was set at 5%. All statistical analyses were performed in RStudio (ver. 0.97.551, R Foundation for Statistical Computing).
Results and Discussion
The Brix, pH, and titratable acidity (TA) of the grape musts were relatively uniform among treatments, with some minor differences (Table 3). Although the Brix of the optically sorted hand- and machine-harvested treatments were significantly lower than in other treatments, the difference was only 0.3 Brix from the treatment with the highest level. The lower Brix in these treatments could have resulted from removal of overripe and raisin-like berries, both of which have high sugar, from the product stream by the optical sorter. There were also significant differences in pH of the musts, but those differences did not exceed 0.1 pH units. No impact on TA due to sorting was observed, potentially because the grapes had mostly uniform ripeness with no visible unripe berries. Although significant, the differences in must chemistry were small from a practical perspective and likely had minimal or no impact on wine sensory character or future reaction chemistry. All 18 fermentations proceeded consistently and achieved dryness within six days. The wine chemical compositions at bottling are shown (Table 4). There were small but significant differences among treatments in all measured chemical parameters. In general, differences in ethanol content were driven by sugar content differences in the grapes following treatment. Discrepancies in grape sugar and final ethanol content in wines are potentially due to larger sugar differences due to soak up then reflected in the Brix measurements of the grapes at harvest. In general, the machine-harvested treatments resulted in wines with lower TA, a possible result of more potassium leaching from the skins and combining with tartaric acid to precipitate as potassium. This was observed during sensory evaluations of the wines. Wine pH values were similar to those seen for the respective grapes. All other parameter differences were deemed insignificant in respect to potential sensory impacts.
Brix, pH, and titratable acidity of must for all treatments. Treatments sharing a common letter do not differ significantly at p ≤ 0.05 (n = 3).
Chemical composition of all wine treatments at bottling. Treatments sharing a common letter do not differ significantly at p ≤ 0.05 (n = 3).
Grape phenolics
The total phenolics and tannin concentrations in grape samples were not significantly different among treatments (Table 5). This is not particularly surprising, as one would not expect different harvesting and sorting methods to significantly alter the chemical composition of grapes harvested from the same vineyard. Since the grapes were lightly cleaned during preparation of grape extracts, no MOG that may have led to treatments having different phenolic concentrations were included in the samples. This could partially explain why differences in phenolic concentration were found in the finished wines (Table 5) but not in the grape samples. Additionally, grape samples were fully extracted while treatment effects, which affected berry intactness, were carried through into the wine due to whole berry fermentations.
Total phenol, tannin, and anthocyanin content of grape samples (mg/g berry), wine at bottling (mg/L), and wine after three months of aging (mg/L) as determined by the Adams-Harbertson assay. Treatments sharing common letters within a column and time period do not differ significantly at p ≤ 0.05 (n = 9).
Analysis by RP-HPLC showed that the grapes contained similar concentrations of (+)-catechin and (−)-epicatechin among treatments except for the unsorted, hand-picked treatment, which had significantly lower concentrations (Table 6). The grapes from the unsorted hand-picked treatment also showed considerably lower anthocyanin concentration compared to the other treatments (Tables 5 and 6). It is possible that the lower concentrations found in that treatment are due to variations in the skin to flesh ratio of the berries used to prepare the grape extracts for analysis, as discussed in the following paragraph on anthocyanin concentrations. This could also be due to lower extractability from the skins due to a higher percentage of raisin-like berries. Tannin levels determined by RP-HPLC were much lower than those determined by the Adams-Harbertson assay, which is a protein precipitation method. This is due to the fact that RP-HPLC underestimates tannin levels as a result of the slow bleeding of unresolved tannins on the baseline (Peng et al. 2002). Previous comparisons found that Adams-Harbertson data correlates best to perceived astringency in wines and is thus the preferred method to determine tannin levels in wine, especially when correlating with sensory analysis (Kennedy et al. 2006). The anthocyanin levels of the grape samples measured using the Adams-Harbertson assay (Table 5) were comparable to values from RP-HPLC analysis (Table 6). Looking only at the unsorted treatments, the two mechanically-harvested lots had greater anthocyanin levels than the hand-picked treatment by an average factor of two. This large difference may be the result of a greater skin-to-flesh ratio in the machine-harvested grapes caused by damaged berries losing juice and pulp. Since anthocyanins are derived solely from the skin in Pinot noir grapes (Boulton 2001, Ribéreau-Gayon et al. 2006), a greater skin-to-flesh ratio from damaged berries would lead to artificially elevated anthocyanin concentrations after back-calculating to mg/g berry units. These differences were not reflected in the total phenolics and tannin measurements (Table 5), which could indicate solubility limitations, as found when the impact of saignée (juice run-off) was investigated using different fruit sources (Gawel et al. 2001).
Phenolic profiles of grape samples (mg/g berry), wine at bottling (mg/L), and wine after three months of aging (mg/L) as determined by RP-HPLC. Treatments sharing common letters within a column and time period do not differ significantly at p ≤ 0.05 (n = 3).
In hand-harvested grapes, optical sorting significantly (p ≤ 0.05) increased anthocyanin concentration (Table 6). A possible explanation for this is that the optical sorter removed raisin-like and sun-damaged berries that provide poor anthocyanin extraction, increasing the concentration in the optically-sorted sample. Optical sorting had a smaller and non-significant impact on anthocyanin concentrations in mechanically-harvested treatments due to slightly higher variability in anthocyanin concentrations within fermentation replicates. While there was a significant difference in anthocyanin concentrations between the unsorted hand-harvested and both unsorted mechanically-harvested treatments, optical sorting minimized variability in the grapes and the sorted treatments were not significantly different. There were no significant differences in quercetin-glycoside concentrations among the different grape samples and only small differences in hydroxycinnamic acid concentrations (data not shown).
HS-SPME-GC-MS analysis of grape volatiles
Statistical analysis of the GC-MS data for grapes indicated that treatment replicates were not significantly different (p ≤ 0.05), allowing the treatment replicates to be combined to assess compound differences attributable to treatment alone. Of the 44 volatile compounds analyzed in the grape samples, 22 compounds had significantly different levels among treatments: ethyl acetate, ethyl hexanoate, nerol oxide, benzyl alcohol, isoamyl acetate, cis-3-hexen-1-ol, trans-2-hexen-1-ol, trans-3-hexen-1-ol, cis-2-hexen-1-ol, β-citronellol, ethyl vanillate, ethyl 2-methylbutyrate, geraniol, unidentified sesquiterpenes I and II, γ-nonalactone, nerol, ethyl octanoate, β-linalool, β-myrcene, damascenone, and α-terpinene. A PCA of the significant compounds shows the variables and individual factor maps (Figure 1). The PCA plot grouped treatments with the same harvest method near one another, indicating that harvest method had a greater influence on volatile profile in the grapes than sorting. Specifically, the hand-picked grapes had higher concentrations of the terpenes and alcohols located in the right quadrants of the loadings plot.
Loadings (A) and score (B) plots of a PCA of the volatile compounds that differed significantly (p ≤ 0.05) among grape samples as determined by HS-SPME-GC-MS analysis.
The PCA loadings plot illustrates that separation of treatments is also driven by β-linalool, β-myrcene, β-damascenone, and α-terpinene. These compounds, which are characterized by floral, spice, and perfume aromas, had greater concentrations in the machine-harvested treatments. The higher concentrations of these compounds in the mechanically-harvested treatments may have resulted from glycosidic hydrolysis of their non-volatile precursors. To a great extent, aromatic compounds in grapes are glycosidically bound, serving as an important reserve of aroma in wine (Williams 1993). The grape-derived glycosidase enzymes capable of liberating aroma compounds are located in the juice and pulp of the berry (Aryan et al. 1987). Accordingly, the berry damage incurred during mechanical harvesting, which disrupts the compartmentalized flesh fraction of grapes, can release glycosidase enzymes which may have led to the greater concentrations of aroma compounds in machine-harvested treatments. Another possible explanation is induction of synthesis as a wounding response to berry damage (Niinemets et al. 2013, Rodríquez et al. 2013). Further study is needed to determine why these particular terpenes are present in higher concentrations in the non-hand-harvested treatments.
Color and phenolic measurements in wines
The hue of all wines was measured at bottling and three months post bottling (Figure 2). The hand-harvested and Selectiv’ treatments had similar hue patterns: hue increased with optical sorting at both times. Conversely, the machine-harvested treatments had decreased hue with optical sorting. At three months after bottling, the three harvest methods produced wines with significantly different (p ≤ 0.05) hues if the berries were unsorted, with the machine-harvested treatment having the greatest value. However, optically sorting the fruit minimized the hue differences so that the values were no longer significantly different. This shows that by removing undesirable grapes and MOG, the optical sorter produced wines of similar hue regardless of harvest method.
Hue of wines (absorbance at 420nm/520 nm) at 0 and at three months after bottling. Treatments sharing common letters within a time period do not differ significantly at p ≤ 0.05 (n = 3).
In general, sorting did not affect color density within harvest treatment except for Selectiv’, which decreased color density with optical sorting (2.36 ± 0.06 HHNS, 2.18 ± 0.08 HHVS, 2.63 ± 0.06 PSNS, 2.31 ± 0.24 PSVS, 2.15 ± 0.07 MHNS, 2.08 ± 0.03 MHVS). After three months in bottle, however, the differences were no longer significant and color densities across treatments were similar.
Optical sorting decreased total phenolics in the wines after three months in bottle, except for in the machine-harvested treatment (Table 5). This result disagrees with a previous study that found a general increase in total phenolic levels in wine made from optically-sorted fruit compared to an unsorted counterpart (Falconer et al. 2006). In general, optical sorting also decreased levels of gallic acid, (+)-catechin, (−)-epicatechin, and tannin in wines at bottling and after three months of aging (Table 6). Some MOG contain high levels of phenolic compounds and it has been shown previously that wines made with the addition of MOG have greater total phenolic content than the control (Huang et al. 1988). It is possible that the general decrease in phenolic compounds observed in the optically-sorted treatments was due to the removal of MOG by the sorter. At both times, the unsorted Selectiv’ treatment had the greatest concentration of total phenolics, gallic acid, catechin, epicatechin, and tannin (Table 6). With the Selectiv’ process on-board, grapes harvested with the Selectiv’ harvester experienced an additional physical process that potentially could have increased berry damage. Any damage occurring during harvest would lead to greater extraction during subsequent transport of the fruit to the winery and during fermentation, since all treatments underwent whole-berry fermentation. Merlot fermentations with different percentages of crushed fruit (0, 25, 50, and 100%) had increased tannin in the final wine as the percentage of crushed fruit increased, with a maximum at 75% crushed fruit (Cerpa-Calderón and Kennedy 2008). This study supports the assumption that a higher percentage of crushed or damaged fruit during fermentation will result in more extraction of phenolics. On harvest day, the mechanically-harvested fruit showed visible damage and the half-ton bins in which the grapes were transported contained a large amount of juice that had leeched from ruptured berries. The Selectiv’ treatment that was optically sorted had phenolic concentrations more consistent with the other treatments, suggesting that the sorter may have effectively removed damaged berries from the process stream, thereby limiting extraction.
The total anthocyanin levels in the different wine treatments at bottling and after three months of aging show a consistent decrease in concentration with bottle aging (Tables 5 and 6), which is expected as anthocyanins react with wine components to form mainly polymeric pigments (Fulcrand et al. 2006), although precipitation and/or oxidation reactions could also contribute to these decreases. Similar to the other phenolic compounds, the unsorted Selectiv’ treatment generally had the highest concentration of anthocyanins. This treatment also had the highest hydroxycinnamic acid and quercetin glycoside concentrations (data not shown). The greater extraction caused by fewer intact berries during fermentation, as discussed earlier, could also explain the higher concentrations in the unsorted Selectiv’ treatment (Cerpa-Calderón and Kennedy 2008). Again, if all treatments had been crushed rather than undergoing whole berry fermentation, these differences may have been eliminated or decreased. For the hand- and machine-harvested treatments, sorting did not lead to large differences in anthocyanin concentrations.
Optical sorting significantly decreased (p ≤ 0.05) polymeric pigment levels in the hand- and machine-harvested treatments at bottling (Table 6). After three months in bottle, however, the differences in polymeric pigment concentrations had decreased and were no longer significant. All treatments also had increased polymeric pigment levels after three months aging, which is expected as polymeric pigments are formed with age as anthocyanins bind to flavanols (Fulcrand et al. 2006).
HS-SPME-GC-MS analysis of wine volatiles
Of the 51 aroma compounds investigated, 45 and 40 were significantly different among treatments at zero and three months after bottling, respectively. This indicates that differences in aroma compound concentrations decreased over time and that the wines became more similar with age. Many commercially produced red wines are aged for considerably longer than three months before they are bottled and released. If the trend of decreasing aromatic differences with time seen here were to continue, the volatile profiles of the wines after a realistic aging period could potentially be even more similar across treatments. However, to test this hypothesis, more analyses over time are required.
A PCA of the scaled data for significant compounds at three months in bottle indicates that the hand-harvested treatments had similar volatile profiles, as they are near one another spatially (Figure 3). The compounds are fairly evenly distributed throughout the loadings plot, indicating that no compounds in particular are driving separation among treatments. The mechanically-harvested treatments group together as well, except for the optically sorted Selectiv’ treatment, which is isolated.
Score (A) and loadings (B) plots of a principle component analysis (PCA) of scaled data of significantly different (p ≤ 0.05) volatile compounds analyzed by GC-MS in wines after three months in bottle (n = 9).
Descriptive sensory analysis
MANOVA revealed significant differences among wine treatments (p ≤ 0.05) for the sensory attributes investigated. A series of ANOVAs were subsequently performed, revealing that only tropical fruit and hue saturation were significantly different among treatments out of the 18 attributes used to analyze the wines. For these two attributes, there was a significant judge-by-treatment and treatment-by-replicate interaction for hue saturation (data not shown). However, the F-ratios with the interaction terms in the denominator were highly significant, indicating that the main effect treatment was more important than the individual interactions, allowing its continued discussion. The attribute means and Fisher’s LSD for the significant attributes are shown (Table 7). The unsorted hand-picked and unsorted Selectiv’ treatments had significantly more tropical fruit aroma than the other mechanically-harvested treatments. Interestingly, this is in disagreement with a previous study on optical berry sorting that found greater tropical fruit character in wines made from sorted fruit (Falconer et al. 2006). It should also be noted that although significantly different among treatments, tropical fruit was not a prevalent characteristic in general, with the highest rating at only 2.47 on a 10-point scale. This is not particularly surprising, as tropical fruit is a relatively uncommon descriptor for Pinot noir and most other red wines.
Overall means and Fisher’s least significant difference (LSD) for the descriptive analysis attributes (n = 13) rated in the wines three months after bottling. Within a row, means sharing a common letter are not significantly different (p ≤ 0.05). Rating scale: 0 (low intensity) to 10 (high intensity).
For each harvest treatment, optical sorting significantly decreased hue saturation. This result is consistent with what would be expected in the context of this study, since the grapes were not crushed before fermentation. With the removal of damaged berries, optically-sorted treatments have more intact fruit, thus limiting extraction of phenolics during fermentation and leading to wines of lighter color. These results are consistent with the UV-vis data that showed small decreases in color density with optical sorting (data not shown), though those differences were not statistically significant.
Since only two attributes differed significantly, with one of them a visual assessment, it is safe to conclude that wines made from the different treatments were quite similar in taste and flavor profile. Within the context of this study, then, it is apparent that mechanically-harvested fruit did not produce inferior wines, as they were mostly indistinguishable from wines made from hand-picked grapes. Similarly, optical sorting did not increase wine quality as the wines were seen as very similar by the sensory panel. Although chemical analyses showed significant differences in several measures of wine composition, these differences mostly did not translate to sensory differences detectable by panelists.
A PLSR related the significantly different GC-MS volatile compounds after three months in bottle to significant attributes in the corresponding DA (Figure 4). The first two dimensions explain 56% of the variance. Tropical fruit and hue saturation, the only significant DA attributes, grouped together in the top right quadrant nearest the unsorted handpicked treatment, with the other two unsorted treatments in the same quadrant. Some compounds in the vicinity that may have contributed tropical fruit character to those treatments are β-citronellol and ethyl acetate, both of which can be perceived as fruity at low concentrations, geraniol, which is found in many essential oils and is used in pineapple and grapefruit flavorings, and nerol, a monoterpene with a fresh scent that is found in lemongrass and hops (Fahlbusch et al. 2003).
Overlaid score and correlations plot of partial least squares regression analysis between significant (p ≤ 0.05) volatile compounds in wines analyzed by GC-MS after three months in bottle and significant attributes from the corresponding descriptive sensory analysis.
Conclusion
A primary objective of this study was to investigate possible synergistic effects, if any, of using mechanical harvesting in conjunction with optical berry sorting on grape and wine composition. In some instances, such as anthocyanins in grape samples and flavan-3-ols, tannins, and hue values in wines, the different harvest treatments did result in significantly different values. However, by implementing optical berry sorting, these differences were either reduced or eliminated altogether. Under the conditions specific to the current study, it is fair to conclude that optical sorting successfully diminished the differences that arose from harvesting by machine. Some chemical differences that persisted into the finished wines, such as anthocyanin concentration and total phenolic concentration, were likely exaggerated by the whole-berry fermentations employed by this study. By allowing differences in berry condition (presumably caused by treatment) to persist throughout fermentation, this technique allowed for differences stemming from treatment method to be seen more clearly. However, since it is common practice in red wine production for the fruit to be crushed before fermentation, the variable phenolic extraction seen among treatments in this study might have been eliminated had more typical processing occurred. Although significant chemical differences were found among treatments, those differences generally did not result in wines that were distinguishable using aroma, taste, and mouthfeel attributes. Only two of the 18 wine sensory attributes were significantly different among treatments, indicating that the wines were very similar in character overall. Pinot noir was specifically chosen as a potentially more sensitive variety in which to investigate the impact of mechanical harvesting due to its lower phenol content compared to varieties such as Cabernet Sauvignon and Merlot. Although this study examined only one season, our findings concur with the few studies completed on other varieties and with many anecdotal reports that seasonal and site impact are minimal, if not negligible. An important exception is years with high rot or mold, when mechanical harvesting is not recommended. As mechanical harvesting and optical sorting become more commonplace in wine production, future studies of a similar nature using other grape varieties are warranted.
Acknowledgments
With considerable gratitude, the authors thank the Napa Valley Vintners, Walsh Vineyard Management, Wine Spectator, the Leon D. Adams Research Scholarship, the Brotherhood of the Knights of the Vine, the Richard and Saralee Kunde Scholarship Fund, the Horace O Lanza Scholarship, Silverado Premium Properties, Pellenc America, Inc., and Bucher Vaslin for financial and material support.
- Received December 2014.
- Revision received August 2015.
- Revision received April 2016.
- Accepted April 2016.
- Published online September 2016
- ©2016 by the American Society for Enology and Viticulture
Literature Cited
Vol 67 Issue 4
Jump to section
Related Articles
Cited By...
More from this TOC section
Similar Articles
