Abstract
Rachis, raisins, rotten, and sun-burnt berries were removed from hand-harvested Chardonnay by automated color sorting. Rejected material comprised 4% (w/w) of the crop. Wine was made from sorted and unsorted fruit and was subjected to chemical and sensory analysis. Chemical analysis found that the sorted wine had a higher total phenolic level, pH, and residual sugar content. Sensory analysis showed sorted wine significantly differed in two attributes: higher tropical fruit aroma and higher sweetness. The two wines could not be strongly differentiated in other sensory characteristics, showing great similarity in palate attributes.
Harvested winegrapes are heterogeneous by nature and can include substandard berries (raisins, rotten, and sunburnt), material other than grape (MOG), and prime grade berries. The contribution to finished wine aroma and flavor from MOG is widely accepted but is often based on experience, rather than scientific evidence. Purchasing agreements often incorporate maximum levels for MOG in order to reward grape growers for improved quality fruit (Allan 2004). Substandard berries can also affect finished wine quality. Berries with powdery mildew infection can impact finished wine quality (Calonnec et al. 2004), with as few as 1 to 5% of infected bunches resulting in detectable differences in finished Chardonnay wine composition and sensory properties (Stummer et al. 2005).
Harvest methods impact the composition of harvested winegrape crops. Hand-harvested winegrapes usually contain rachis but are substantially freer of other forms of MOG compared with mechanically harvested fruit. Hand picking allows for selectivity, thus minimizing harvesting of substandard berries. Mechanical harvesting provides less control over amounts of substandard berries, MOG, and foreign materials because of the nonselective nature of the process. Blowers can be used to eject leaves and other light foreign materials from the mechanically harvested product (Boulton et al. 1996). Mechanical harvesting provides some advantages to grapegrowers, especially when economical labor is not available (Clary et al. 1990). It also allows rapid harvesting when a crop reaches its peak and is amenable to operation during cool conditions.
Postharvest sorting is one way to eliminate unwanted material from grapes, particularly from fruit used to produce high-value wines. Whole bunches can be sorted to eliminate rotten fruit and large pieces of MOG. Single grapes separated from the rachis in a destemmer can be sorted to eliminate remaining raisins and MOG.
Automated sorting using optical properties of light reflected from the surface of the fruit to differentiate between good and bad material is one alternative to hand sorting (Abbott 1999). Automated sorting can theoretically inspect every berry to a degree not practical by hand and provide a more consistent and efficient process. Automated sorting technology has been developed that combines illumination of the product, capture of the reflected light by charge-coupled device camera, computer analysis of the image, and an ejection mechanism to eliminated unwanted materials. Image-processing techniques used for food-quality evaluation are based on shape, color, and texture and have been applied to a wide variety of foods, including fruit, vegetables, grains, and meats (Du and Sun 2004). Commercial instruments that use image-processing techniques for sorting food products can be adapted for winegrapes. Preliminary studies indicated that automated sorting was compatible for use with both hand and mechanically harvested grapes and successfully eliminated MOG, raisins, and brown berries from varieties of white grapes, and MOG and green berries from varieties of red grapes. It was unable to distinguish rotten berries from varieties of red grapes, however (Falconer and Hart 2005).
The first step in incorporating a new operation such as automated sorting into a commercial process is to assess whether its use has any negative impact on product quality. In the trial described here, automated sorting was applied to a hand-harvested Chardonnay crop with minimal defects. The MOG component consisted of the rachis and a few dried leaves. Substandard berries were comprised of a small number of rotten berries, raisins, and sun-burnt berries. Results demonstrated the impact the automated sorting operation had on processing typically clean, good-quality hand-harvested fruit.
Materials and Methods
Grape source.
Hand-picked Chardonnay grapes were sourced from Seppelt Vineyard, Heathcote, central Victoria, Australia (Figure 1⇓). Grapes were picked early in the morning to minimize high temperatures and had a target Brix of 22.9. Potassium metabisulfite solution was added to each bin to attain 50 mg/L total sulfur dioxide levels (TSO2). Upon arrival at the winery, the fruit was split into two batches of ~1 tonne (metric ton).
Sorting operation and must preparation.
The configuration used in this trial was designed to process half-tonne bins of grapes. One tonne of the grapes was poured into a hopper above a destemmer (Vaslin Bucher Delta E2; Chalonnes sur Loire, France) in which bunches were broken up and large rachis and canes removed. The flow rate through the destemmer was controlled by the speed of the auger that fed the destemmer so that ~6 tonnes per hour were being processed. Grapes were then transferred by a 32-cm wide conveyer to a vibrational separator (Constructie Bruynooghe, Staden, Belgium), which spread the grapes into an even stream that fed into the color sorter (Sortex, London, UK). Grapes entered the sorter through a power slide, below which was a well-illuminated open zone where the image of each berry and item was captured. These images were used to distinguish between acceptable and reject material. Immediately below the open zone was a bank of 160 air ejectors that expelled defective material from the product stream using short blasts of air of 7 bar. The sorter was programmed to eliminate brown berries by color, while rachis, leaves, and petioles were removed by shape. Accepted berries were directed into half-tonne bins, and the rejected material was removed by conveyer. Juice collected from the destemmer, vibrator, and sorter was pumped into a collecting tank.
Unsorted fruit was pressed as whole bunches in an air-bag press (EHP 4000 Franz Scharfenberger, Bad Dürkheim, Germany). Sorted fruit was pressed with the same equipment. Acidity was adjusted in all juice batches to a titratable acidity (TA) not exceeding 6.5 g/L after initial analysis. If pH was greater than 3.5, then TA was readjusted to attain a pH of 3.5. Cold settling occurred for 18 hr and then juices were racked. Novozyme Ultrazyme CPL (Bags-vaerd, Denmark) was added to the juice (0.01 mL/L). A second cold settling for 18 hr followed by racking and filtration through a 10-μm cut-off filter was used to clarify the juice.
Fermentation.
Wines were made in 20-L demijohns in triplicate. Yeast (EC1118, Lallemand, Montreal, Canada) was added at 200 mg/L after rehydration according to manufacturer’s recommendations. Fermentation was conducted between 15 and 18°C. When residual sugar reached <2 g/L, the ferment was chilled to <5°C and 0.5 g/L bentonite and 60 mg/L SO2 were added. Copper sulfate was added to eliminate any potentially reductive characteristics. After fermentation the demijohns were placed in a cool room at <5°C for 10 to 15 days to settle and allow minor tartrate precipitation. Once settled, wines were racked under gas to sterile vessels of appropriate size and free sulfur dioxide (FSO2) was adjusted to 20 to 25 mg/L and TA to ~6.0 g/L. No filtration or malolactic fermentation of the wine occurred and no oak adjuncts were added. Wine was then bottled into 750-mL bottles under gas cover and screwcap closures were applied. Chemical and sensory analyses were done 12 weeks after bottling.
Chemical analysis.
All wines were analyzed for free and total SO2, alcohol, residual sugar, pH, TA, volatile acidity, OD420, OD520, OD280, and metal ions by Analytical Services of the Australian Wine Research Institute (Urrbae, South Australia) using standard analytical techniques (Iland et al. 2004). Phenolic analyses were conducted by ETS Laboratories (St. Helena, CA) using high-performance liquid chromatography.
Sensory analysis methodology.
All wines were stored in a wine cellar at 15°C for a minimum of one month before sensory assessment. Sensory tests were conducted in a purpose-built sensory room using nine individual booths. Data acquisition and analysis for all sensory evaluation was carried out using Fizz data collection software (version 2.0, Biosystèmes, Couternon, France).
Difference testing.
A duo-trio difference test was performed for aroma and flavor with a panel of 25 trained judges using the balanced reference duo-trio method (Meilgaard et al. 1999). All samples were presented for assessment in 30-mL aliquots in three-digit-coded, covered ISO standard tasting glasses under orange sodium lights at 22 to 23°C. The judges were staff of Provisor Pty Ltd and the University of Adelaide and all had experience in winetasting and difference tests. Each judge was presented with a reference sample first, followed by two samples, one of which was identical to the reference that the judge was asked to identify. The test samples were blends of the three fermentation replicates for each wine treatment. There was one test comparison made between a sorted wine sample and an unsorted wine sample.
Descriptive analysis.
A panel of 13 judges (10 female, 3 male, ages 35 to 60) were convened for this trial, all staff of Provisor employed as part-time sensory judges experienced in descriptive analysis studies. During two training sessions, the judges generated, by consensus, a list of attributes (11 aroma and 9 by mouth; taste, flavor, and mouthfeel) considered necessary to differentiate and describe the wines. Judges were trained to recognize these sensory attributes by developing aroma reference standards that were presented at each training session, with judges requested to comment on their suitability to define the terms rated in the wines. Reference mouthfeel standards for acidity, sweetness, drying, and viscosity were presented at the first training session and a final set of attributes and definitions were determined (Table 1⇓). Two practice rating sessions were then held under the same conditions as subsequent formal sessions to calibrate their intensity-measurement framework.
Samples were assessed in a formal session and served at 22 to 23°C under orange lights to minimize visual cues. Wines were presented in triplicate in random order to the judges in three-digit-coded, covered ISO standard tasting glasses (25 mL). The judges rated each of the attributes on an unstructured 15-cm line scale, with anchors of “low” and “high” placed at 1 cm and 14 cm, respectively.
The descriptive data for each attribute was assessed using a nested analysis of variance (ANOVA) based on a balanced incomplete block design, testing for the effects of treatment. Fisher’s least significant difference (LSD) means comparison test (p = 0.05) was also performed.
Results
The Chardonnay grapes were hand harvested and supplied in half-tonne bins. The fruit was of good quality containing small numbers of rotten fruit, sun-burnt fruit, and raisins (Figure 1⇑). There was little MOG other than the rachis component of the grape bunches and a few leaves. There was no sign of physical damage to the grapes during transport. Two half-tonne bins were tipped by forklift into the hopper above the destemmer.
The color sorter was adjusted to remove material based on both color and shape. The color map was set to eliminate any berries with a high proportion of brown color. The shape detector was set to eliminate material that matched the dimensions of rachis and petioles. The accepted grapes were predominantly green, clean berries with little MOG. The reject material contained approximately 25% rachis, 25% raisins, 50% sun-burnt, and a few good berries and dried leaves (Figure 2⇓). The reject material constituted 4% (w/w) of the total material processed.
Reject material was pressed in a small lab-scale pneumatic press and was only used for visual comparison. Juice prepared from the rejected material had a brown color compared to juice pressed from sorted and unsorted grapes. Chemical analysis of the juice from unsorted and sorted (accepted) grapes indicated that they were very similar (Table 2⇓).
Chemical analysis of the finished wines showed minimal differences between the two wines (Table 3⇓). There were significant differences between pH, total phenolics measured as the optical density at 280 nm (OD 280 nm), and glucose-fructose analyses. The wine made from sorted grapes had a higher pH and OD 280 nm reading with higher residual sugar levels.
The phenolic analysis assay yielded results typical for Chardonnay grapes (Table 4⇓). The wines made from unsorted and sorted grapes could only be differentiated by the concentration of caffeic acid and grape reaction product (GRP).
Comparison of the sensory difference testing results alone was statistically inconclusive, but indicated that the sorted and unsorted wines could be differentiated by aroma and flavor (p = 0.105). Seventeen of the 25 judges who initially gave correct difference test responses commented that the sorted wine had a greater floral, ripe tropical aroma with a sweeter taste. The judges rated these differences as small to medium.
The mean intensity scores for each aroma and flavor attribute were plotted (Figures 3⇓ and 4⇓). An analysis of variance of the sensory descriptive analysis data for the effects of judge, wine, and replicate for the two wines (Table 5⇓) showed that wine sensory properties were significantly different between the two treatment groups for two attributes: tropical fruit aroma and sweetness. Wine made from sorted grapes was higher in tropical fruit aroma with a sweeter taste. Sensory evaluation of perceived acidity was inconclusive as the p value was exactly 0.05, although the higher intensity of perceived acidity in wine from unsorted grapes coincides with a lower pH value.
Discussion
Results showed that automated sorting of Chardonnay winegrapes can significantly alter chemical and sensory characteristics in the resultant wine. Chemical analysis identified the wine produced from sorted grapes as being higher in total phenolics (OD 280 nm), pH, and residual sugar compared with the wine from unsorted grapes. Sensory analysis indicated the wine made from the sorted grapes was perceived to be sweeter, possibly lower in acidity, and higher in tropical aroma. This would generally be considered to be positive. However, residual sugar itself is derived from incomplete sugar utilization during the fermentation process and cannot be directly linked to the sorting process. This phenomenon is difficult to explain by the absence of MOG, sun-burnt fruit, rotten fruit, or raisins unless these materials provide nutrients that enable greater utilization of the available sugar. Given that the sensory threshold for residual sugar is well above the levels recorded in this study, it is probable that the judges perceived the wine to be sweeter because of differences in acidity rather than a difference in residual sugar. The two wines showed similar palate profiles and could not be differentiated in any sensory characteristics other than sweetness, acidity, and tropical fruit aroma.
This study was not designed to identify which step in the sorting process was responsible for the changes in sensory attributes and chemical composition of the finished wine. The sorted fruit was subjected to destemming, vibrational separation, and color sorting before collection in half-tonne bins and pressing. The unsorted fruit was pressed as whole bunches. One difference between the sorted and unsorted grapes is the separation of the rachis from the berry during the destemming operation, which initiates some berry maceration and juicing with an opportunity for oxidation to occur. The physical nature of the destemming and spreading processes may also impart a low level of trauma onto the berry, while the pressing operation will also differ between single grapes and whole bunches because of the inclusion of the rachis in the whole bunches. Extraction of higher levels of phenolic compounds and the chemical compounds responsible for the release of tropical fruit aroma could well be a result of maceration or differences in the pressing operation. Berry trauma during processing could also be a component. Obvious modifications to the process could help reduce the amount of maceration. The receiving bin could be replaced by a small hopper with direct transfer of sorted berries to the press or the machine-harvested fruit could bypass the destemming process altogether.
A preliminary study with hand- and mechanically harvested winegrapes (Falconer and Hart 2005) demonstrated that the color-sorting operation is not limited to hand-harvested Chardonnay. It can also be applied to mechanically harvested fruit and used to remove green or raisin berries from red grape varieties such as Zinfandel or rotten brown berries and raisins from white grape varieties. There is evidence that substandard fruit such as grapes infected with powdery mildew (Calonnec et al. 2004, Stummer et al. 2003) and lactic acid bacteria (Costello et al. 2001) negatively impact finished wine quality.
Automated sorting does have limitations. Mechanically harvested winegrapes that are badly damaged or pulped are difficult to sort effectively, which limits the usefulness of automated sorting for fragile grape varieties and fruit subjected to damage or dejuicing during transportation. Automated sorting is an economically feasible operation for medium-sized wineries with sorting running at capacity (Falconer and Hart 2005). In years with healthy crops, this automated sorting is only valuable if an improvement in the finished wine quality can be identified. In years with specific problems such as rot, unripe fruit, or bird peck, automated sorting may benefit winemakers.
While automated sorting has obvious potential for the winemaking process, it could also be a useful tool for enologists to study the impact that materials such as petioles, raisins, rotten and sun-burnt berries have on finished wine attributes. Ejected material could be fed back into sorted “clean” fruit and their impact on the finished wine’s sensory properties determined. Additional sensory attributes such as mushroom and earthy aromas should also be incorporated in such studies as they are likely to be relevant measures in wine quality (Stummer et al. 2005). The amount of MOG or substandard berries required for the defect to be detectable in the finished wine could be identified, allowing for meaningful limits for MOG and substandard berries in purchasing agreements.
Conclusion
Results showed that automated color sorting of Chardonnay winegrapes is a feasible operation and has obvious application for wineries already using hand sorting to remove undesired materials from incoming grapes. The sorting operation had a detectable impact on a minority of the chemical and sensory characteristics in the finished wine. Higher tropical aromas would generally be considered favorable; however, higher phenolic levels could be considered either positive or negative depending on intended wine style. Modification to the sorting process may help reduce maceration and phenolic levels in the resulting wine. Crops with obvious faults such as raisin or green fruit, bird pecking, or other identifiable problems may benefit from automated sorting. Ultrapremium wine producers who currently use hand sorting to eliminate defective grape berries and MOG may also consider automated sorting as an economically viable alternative.
Footnotes
Acknowledgments: The authors acknowledge Sortex Limited, London, for financing and providing the equipment for this study and Fosters Wine Estates, Southbank, Australia for providing the grapes and facilities.
The authors also thank David McCambridge and Ben Deefholts of Sortex, Norm Woodhams of Seppelt Wines, Vincent O’Brien and Chris Day of Provisor, and Andrew Fleming and the staff of Coldstream Hills Winery for their assistance during this work.
- Received February 2006.
- Revision received May 2006.
- Copyright © 2006 by the American Society for Enology and Viticulture