Abstract
Relationships among sensory attributes, compositional measures, and wine quality of Shiraz grapes and wines were evaluated for two seasons, 2009 to 2010 and 2010 to 2011. The sensory profiles of berries and wines were evaluated by descriptive analysis and wine quality was assessed by an expert panel. In this study, berry sensory attributes alone were better predictors of wine sensory and compositional variables than the combination of berry sensory and compositional variables. Partial least squares regression analysis and Pearson’s correlation revealed a negative relationship between seed bitterness and wine savory spice flavor in both seasons. In 2011, pulp detachment from the skin correlated with wine sensory attributes such as rim color, fresh dark berry flavor, savory spice flavor, and wine quality score. Correlations among wine sensory attributes, wine pigmented polymers, and wine total tannins were identified in both seasons. These findings are important for grapegrowers and winemakers as they identify berry sensory attributes that may assist as objective measures in predicting final wine style and quality.
- berry sensory assessment
- descriptive analysis
- wine sensory attributes
- seed bitterness
- wine quality score
- savory spice
- pulp detachment
Berry sensory assessment (BSA) is a grape-tasting method used by grape and wine producers to evaluate sensory characteristics of the whole berry and its individual parts (skin, pulp, and seeds). Interest in winegrape sensory evaluation has increased in the last decade, due in part to the increasing familiarity of wine producers and researchers to BSA as a practical method (Rousseau and Delteil 2000, Winter et al. 2004, Olarte Mantilla et al. 2012). Recent research has explored the effect of leaf removal (Lohitnavy et al. 2010), temperature (Sadras et al. 2013), water stress (Bonada et al. 2013), and vineyard location (Le Moigne et al. 2007) on grape sensory attributes as determined by BSA.
In a recent survey, members of the wine industry highlighted that “understanding the relationships in the grape berry-wine continuum” was key knowledge needed to better use BSA (Olarte Mantilla et al. 2012). Winemakers and grapegrowers reported that understanding the relationships of the berry-wine continuum may enable them to predict optimum harvest time for improved wine quality, simplify allocating fruit to low- or high-grade end products, and determine practices to better manage berry flavor development and composition in the vineyard to make wines with desired characteristics.
Some insights into the relationships between BSA and wine sensory characteristics have been provided for Shiraz (Davidson Viticultural Consulting Services 2006), Chardonnay (Rousseau 2001), Grenache (Pozzo Di Borgo and Rousseau 2004), and Shiraz and Merlot (Winter et al. 2004). The work on Shiraz showed that high green character sensory scores for berries and wines were associated with higher yields and under 50% bunch exposure (Davidson Viticultural Consulting Services 2006). Also, berries with higher sensory scores for berry fruit flavor and flavor length were associated with higher scores for jam aroma and body in the corresponding wine. A positive correlation was found between BSA pulp fruit character attributes and wine body for Chardonnay (Rousseau 2001, Winter et al. 2004).
As in the studies mentioned above, the aim of this research was to identify possible relationships between berry and wine sensory attributes and compositional variables in Shiraz grapes and wines. However, unlike earlier research, our data collection encompassed two seasons of objective berry and wine sensory evaluation and used the formal sensory science technique descriptive analysis (DA), in contrast to routine wine industry berry and wine sensory assessment procedures. Additionally, the current research gathered berry and wine compositional measures and determined any existing relationships among berry chemistry and sensory measures and wine composition, sensory, and quality scores using multivariate statistics, including partial least squares (PLS) regression analysis.
Materials and Methods
Vineyard
Shiraz, clone BVRC30, (Vitis vinifera L.) berry samples were sourced from a vineyard at the South Australian Research and Development Institute (SARDI) research station located at Nuriootpa, Barossa Valley, South Australia. The experimental site was planted in 2001 at 1481 vines per ha and trained to a bilateral cordon. Vines were planted at 2.25 m and 3.0 m vine and row spacing, respectively. Vines were spur-pruned to 40 nodes per vine. The soil type is a fine sandy loam A horizon overlying a red-brown B horizon (Northcote et al. 1954). All treatment vines were drip-irrigated and received 1.3 mL/ha (128 mm) in 2010 and 0.6 mL/ha (56 mm) in 2011 of irrigation during the growing season. The irrigation schedule was based on soil moisture probes. Growing season rainfall (September to April) was 267 mm in 2010 and 516 mm in 2011. Long-term annual rainfall for the site is 500 mm (Bureau of Meteorology 2013).
Experimental design
Twelve sampling sites, each comprised of seven vines, were randomly selected across the vineyard. One hundred berries per sampling site were collected randomly on a weekly basis from veraison until harvest for grape maturity measures (total soluble solids (TSS), pH, and titratable acidity (TA)). Berries were collected from different sections of the cluster—top, middle, and bottom—and from both sun-exposed and non-exposed sides. Fruit from each site was harvested at ~25.3 Brix in 2010 and at ~22.8 Brix in 2011. For each of the 12 sites, all clusters from the seven vines were harvested and pooled and a 35-kg sample was collected for sensory evaluation, chemical analysis, and winemaking. From each of the 12 samples, 4.5 kg (~15 clusters) was put aside for sensory and compositional analysis. These clusters were broken into smaller clusters of ~six berries. Two kg were allocated to BSA and were kept at 4°C overnight after collection and then equilibrated at room temperature for two hours prior to being presented in plastic weight boats to panelists for sensory assessment. This process was performed in both seasons.
Winemaking
Wine was made separately from each of the 12 sampling sites. Clusters affected by Botrytis cinerea were excluded (this sorting procedure was performed out of necessity on the 2011 season grapes only due to high disease incidence from higher rainfall). Twenty-five kg of fruit per sampling site was transported to the Hickinbotham Rose-worthy Wine Science Laboratory, the University of Adelaide winemaking facility. Grapes were stored at 4°C overnight and destemmed and crushed the next day using a combined crusher/destemmer (Enoitalia; ENO-15, Cerreto Guidi, Italy) into 30-L plastic fermenters (Ampi, Keysborough, Australia). During crushing, 50 mg/L sulfur dioxide (SO2) was added as a 20% solution of potassium metabisulfite to each of the 12 musts. Prior to yeast inoculation, the pH was adjusted to 3.5 with tartaric acid. The musts were all inoculated on the same day with AWRI 796 yeast (Maurivin, Toowoomba, NSW, Australia) according to manufacturer’s instructions. Diammonium phosphate (0.5 g/L) was added at the time of yeast inoculation when the ferments were between 18 and 20°C. Ferments were hand-plunged three times per day and co-inoculated with malolactic bacteria (Oenococcus oeni) when the ferments were ~14.4 Brix. Once primary fermentation reached ~3.6 Brix, the ferments were pressed with a 130-L capacity bladder press in 2010 (Diemme 130L; Laboratory Press, Lugo, Italy) and a 20-L bladder press in 2011 (Zambelli, Hydro 20, Camisano Vicentino, Italy) into 10-L glass demijohns. When primary fermentation was completed, the ferments were racked off gross lees into 5-L glass vessels, avoiding ullage by adding glass marbles. Additions of SO2 to reach 80 mg/L total SO2, copper sulfate (1.5 mL of 1 mg/L aqueous copper sulfate solution per 750 mL bottle), and tartaric acid (1 g/L if pH was above 3.55) were made to ferments that had completed malolactic fermentation (<0.05 g/L malic acid). The wines were kept at 0°C for three weeks for cold stabilization, then filtered using a pad filter (Colombo-Rover pump and six-pad filter, Polverara, Italy) with 0.8 μm Z6 cellulose filter pads (Ekwip, NSW, Australia) and bottled in 375-mL bottles with metal screwcap closures. The wines were then stored at a constant 15°C for future wine sensory and chemical evaluations.
Sensory evaluation
Sensory evaluation of the twelve berry and wine samples was performed using DA and conducted at the University of Adelaide’s Waite campus sensory facilities. Training sessions were divided into two stages: initial training in the open plan focus group room and final training and evaluation sessions in computerized, individual booths under fluorescent light with a light temperature of 6500°K in the sensory laboratory. Fizz software version 2.47b (Biosystèmes, Couternon, France) was used to collect the sensory ratings and to generate a randomized, balanced sample presentation order for each assessor. One-minute breaks after each sample and five-minute breaks every three samples were enforced and panelists used citrus pectin solution during berry tasting (1 g/L, Sigma-Aldrich Co., St Louis, MO) and plain water crackers (Home Brand; Woolworths Limited, SA, Australia) during wine tasting to alleviate palate fatigue as described (Olarte Mantilla et al. 2013).
BSA: training
Fourteen assessors in 2010 (21 to 72 years old, 11 women and three men) and eleven assessors in 2011 (20 to 32 years old, eight women and three men) were selected for the DA panel based on their experience in sensory evaluation, motivation, and availability. Assessor training for both seasons was conducted as described previously (Olarte Mantilla et al. 2013). The first part of the training was mainly focused on evaluating assessors’ ability to identify basic taste qualities: sweetness, acidity, bitterness, and astringent mouthfeel. Assessors were also presented with taste quality ranking exercises in grape juice as described (Olarte Mantilla et al. 2012) to evaluate their ability to discriminate among intensities. The ranking exercises were also used to establish low and high anchors for tasting qualities in the relevant grape part: skin, pulp, or seed. The second stage of training included evaluating berry sensory attributes of Shiraz berries from the University of Adelaide Coombe vineyard. In this training exercise, assessors familiarized themselves with the tasting procedure and generated new or removed redundant sensory attributes relevant to the Shiraz berry samples.
BSA: final evaluation
Two 2-hr sessions, conducted over two weeks, were required to evaluate the berries from 2010, as all samples did not reach the predetermined TSS at the same time. Berry sensory assessment of 2011 samples was performed in two 2-hr sessions within a week. Assessors evaluated 25 attributes of 12 randomly presented samples twice (each in a separate session) resulting in 24 samples per assessor as described previously (Olarte Mantilla et al. 2013). The attribute color extraction was not evaluated in 2011 because in 2010, comparison with the color evaluating scale was affected by individual assessors’ saliva characteristics.
Wine sensory evaluation: training
Sensory evaluation of wines was performed using DA. Fifteen assessors in 2010 (23 to 44 years old, seven women and eight men) and twelve assessors in 2011 (20 to 38 years old, nine women and three men) were selected for the DA panel based on their experience in sensory evaluation, motivation, and availability. Assessor training in 2010 was conducted in eight 2-hr sessions, while in 2011 it was conducted in four 2-hr sessions. The training sessions involved ranking exercises of taste attributes, astringency, and identification of aroma standards. During the training sessions, the assessors were also presented with several trial wines (four per session in 2010 and six per session in 2011) to develop wine appearance, aroma, flavor, mouthfeel, and aftertaste descriptors. Once assessors had tasted the twelve wine samples, the descriptors were collated and a final list of descriptors was created by consensus. Twenty-three descriptors were generated for 2010 wines; they included two attributes for appearance, eight for aroma, six for mouthfeel, six for flavor, and one for aftertaste (Table 1). Twenty-five attributes were generated to describe the 2011 wines, three for appearance, nine for aroma, six for mouthfeel, six for flavor, and one for aftertaste.
Wine sensory evaluation: final assessment
Final evaluation of the wines from both seasons was conducted in two 2-hr sessions over two weeks. The assessors evaluated 12 samples in duplicate, resulting in 24 samples per assessor (12 samples per session). Assessors were presented with 30-mL samples served in clear INAO (ISO standard) 215-mL tasting glasses covered with a petri dish. Reference standards were available throughout the formal evaluations (Table 1) and assessors revisited them before starting the assessment and were encouraged to do so again during breaks or as required.
Wine quality evaluation: expert panel
In addition to the DA sensory evaluation, a sensory evaluation was conducted to determine wine quality. Wine experts were recruited based on their knowledge and experience and were asked to evaluate the quality of the wines under blind conditions using the Australian wine show 20 point scoring system (Rankine 1990, Dunphy and Lockshin 1998). Eight wine experts participated in the wine quality evaluation in 2010 and 12 in 2011; each was performed within a month after the DA sensory evaluation was completed. Wine quality evaluation was conducted at the University of Adelaide’s Waite campus sensory facility booths under the same room conditions described above. A random balanced presentation design was used to present the wines to the wine experts in duplicate for each season; this resulted in 22 (in 2010, only one bottle was available for two of the samples) and 24 (2011) wines being presented. Thirty mL of each wine sample was presented to the expert assessors served in clear INAO (ISO standard) 215-mL tasting glasses covered with a petri dish and identified with a three-digit code. Paper score cards were provided to record the quality scores.
Berry chemistry
One hundred and fifty berries were randomly sampled from each of the twelve sampling sites at harvest. One hundred berries were crushed and centrifuged, and the supernatant was retained to determine TSS using a digital refractometer (ATAGO Pocket – α, Tokyo, Japan). pH and TA measurements were made using a pH meter and autotitrator (CRISON; CompacT TITRATOR, Barcelona, Spain) as described (Iland et al. 2004). The remaining fifty berries were frozen at −20°C for six months and total anthocyanins and phenolics were determined as described (Iland et al. 2004) (Supplemental Table 1).
Wine chemistry
The wines were subjected to standard chemical analysis (SO2, pH, TA, volatile acidity, alcohol, and residual sugar) at the time of sensory evaluation as described (Iland et al. 2004) (Supplemental Table 1).
High performance liquid chromatography (HPLC) of wine samples
HPLC analysis was used to determine concentrations of anthocyanins, tannins, and pigmented polymers in wines from both seasons as described (Eglinton et al. 2004). Wines were clarified by centrifugation at 8,049 g for four min. Twenty μL wine supernatant was run through a polymeric column (100 Å, 5 μm, 250 × 4.6 mm; Polymer laboratories, Church Stretton, UK) and a guard cartridge of the same material at a flow rate of 1 mL/min. Individual anthocyanins and pigmented polymers were identified at 520 nm and quantified as malvidin-3-glucoside equivalents (M3G eq/L). Different concentrations of malvidin-3-O-glucoside chloride (Extrasynthese, Genay, France) was used to generate the calibration curve. Using the same procedure, tannins were identified at 280 nm and quantified as catechin equivalents (CAT eq/L) from a standard curve prepared with catechin (Sigma-Aldrich) (Supplemental Table 1).
Statistical analysis
A two-way mixed model analysis of variance (ANOVA) with random assessors for each season’s data of BSA and wine sensory scores was executed to identify sensory attributes that were significantly different among the 12 berry samples or the 12 wine samples. Relationships between the sensory attributes of berry and wine samples were determined using the mean scores of each significant attribute from all assessors. Sensory attributes that differed significantly by ANOVA were joined with the compositional data, grouped by modality, and subjected to multiple factor analysis (MFA) to determine their contribution to the sample differences. Three individual MFAs were conducted per season for the relationships between berry sensory attributes and (1) wine sensory attributes, (2) wine compositional variables (Supplemental Table 1), and (3) wine quality score to explain the data variability and to determine whether PLS analysis was appropriate. Three more MFAs were conducted after adding the berry compositional variables. For the variability to be explained, the MFA RV coefficients had to be equal or greater than 0.6 (data not presented). The RV coefficients were sufficiently large to make PLS regression analysis appropriate, except for the dataset combining berry sensory attributes, berry compositional variables, and wine quality scores for 2010.
PLS regression was performed for each season to identify berry sensory attributes and berry compositional variables that could predict (1) wine sensory attributes, (2) wine compositional variables, and (3) wine quality scores. Q2 coefficients were compared to test if the combination of berry sensory attributes and berry compositional measures produced a more reliable prediction model than berry sensory attributes alone. After PLS analysis, sensory attributes that had a variable importance for prediction (VIP) equal or greater than 0.8 were included in the prediction model. Models with a Q2 cumulative value greater than 0.4 were considered suitable predictors of wine sensory attributes, wine compositional measures, or wine quality scores.
Pearson’s correlation was also conducted for each season using all berry and wine sensory attributes and compositional variables, irrespective of whether they were significantly different or not. ANOVA, MFA, PLS, and Pearson’s correlation analyses were performed using the statistical package XLSTAT version 4.02 2012 (Addinsoft SARL, Paris, France).
Results and Discussion
Berry and wine sensory evaluation, composition, and wine quality evaluation
Statistical analyses were conducted for both sessions from the 2011 BSA, but only on data from session two from the 2010 season BSA. The reason for this decision was that the ANOVA of the 2010 season BSA from session one showed significant differences for some assessors in their scoring, resulting in uncertain estimates of the sample mean scores. BSA revealed ten significantly different attributes in 2010 and six in 2011 (p ≤ 0.10, Table 2). Only two attributes, seed bitterness and skin disintegration, were significantly different in both seasons.
Wine sensory evaluation of 12 samples showed 11 significantly different attributes in 2010 and 15 in 2011 (p ≤ 0.10, Table 2). Seven wine sensory attributes were different in both seasons: rim color, body color, fresh red berry aroma, fresh dark berry aroma, alcohol flavor, tannin quantity, and tannin quality. Wine quality scores showed significant differences among the 12 samples in 2011 (p ≤ 0.10), but not in 2010 (Table 2).
Relationships among berry sensory attributes, berry compositional variables, and wine sensory attributes
PLS regression of the 2010 data generated a prediction model where the combination of four berry sensory attributes (pulp fresh fig flavor, pulp prune flavor, skin acidity, and seed bitterness) could predict one wine sensory attribute (wine savory spice flavor) with high Q2 (0.75) and R2 (0.84) coefficients (Table 3). Pulp prune and fresh flavors and seed bitterness were negative predictors for wine savory spice flavor, while skin acidity was a positive predictor (Figure 1A).
PLS regression of berry and wine sensory attributes from 2011 determined that the combination of four berry sensory attributes (pulp acidity, pulp detachment from the skin, seed astringency when crushed, and seed bitterness) could predict three wine sensory attributes (rim color, fresh dark berry flavor, and savory spice flavor). Satisfactory Q2 (>0.40) and R2 (>0.60) values were obtained for the individual prediction of rim color, fresh dark berry flavor, and savory spice flavor (Table 3). The combination of pulp acidity, seed astringency when crushed, and seed bitterness was a negative predictor and pulp detachment from skin was a positive predictor of these three wine sensory attributes (Figure 1B).
In 2010, a similar prediction model was identified when berry sensory attributes were used to predict wine sensory attributes or when both berry sensory and compositional measures were included in the PLS regression (i.e., the Q2 coefficient remained the same) (Table 3). No berry compositional variables could predict wine sensory attributes when used as a combined predictor with berry sensory attributes in 2010.
PLS regression using the 2011 berry sensory attributes and compositional variables determined that the same berry sensory attributes could predict wine sensory attributes as when only berry sensory attributes were used. Pulp acidity, pulp detachment from skin, seed astringency when crushed, seed bitterness, and one berry compositional variable (berry color) were part of the prediction model for wine rim color and wine fresh dark berry flavor. However, the Q2 coefficients of the models for wine rim color and wine fresh dark berry flavor decreased when compositional measures were added to the PLS analysis. It was not possible to obtain a model for wine savory spice flavor with the combination of berry sensory attributes and berry compositional variables (Table 3). Pulp acidity, seed bitterness, and seed astringency when crushed had negative relationships and pulp detachment from the skin and berry color had positive relationships with rim color and fresh dark berry flavor (Figure 1C).
Seed bitterness correlated negatively with wine savory spice flavor in both seasons (r = −0.57, p = 0.05 in 2010; r = −0.71, p = 0.01 in 2011). Wine savory spice attribute was defined by the tasters as a pepper-like sensation. To our knowledge, no previous studies have reported this relationship. However, this wine attribute is associated with rotundone, a compound that is formed in the berry and accumulates exclusively in the exocarp. Its concentration in wine can be up to 10% of the concentration in berries, depending on the extent of skin contact during fermentation (Caputi et al. 2011). Rotundone is usually found in high concentrations in Shiraz wines (Herderich et al. 2012), but may vary in berries sampled across different sections of a vineyard (Scarlett et al. 2014). Rotundone concentration in berries increases as the distance of the berries to the vegetal parts decreases (Zhang et al. 2013). Although Scarlett et al. (2014) did not find a close relationship between berry rotundone concentration and vine vigor, they suggested that rotundone concentration could vary due to vineyard variability and temperature. Understanding pepper sensations in Australian Shiraz wines is of interest, as anecdotally it is an attribute often associated with premium Shiraz wines (Herderich et al. 2012) and can impact consumer preference for wine (Lattey et al. 2010).
Evaluation of seed characteristics is part of a berry sensory analysis protocol (Winter et al. 2004) that is used by Australian wine producers (Olarte Mantilla et al. 2012). However, evaluation of seed bitterness is not included in that protocol. Seed bitterness was negatively associated with both fresh dark berry and pepper-like sensations in wine, which indicates that seed bitterness is an important attribute to include on berry sensory analysis score sheets.
Grape seeds have the greatest bitter intensity of all berry parts (Brossaud et al. 2001), but bitterness in seeds does not necessarily translate to bitterness in wine; the relationship may be confounded by enological impacts. In the current study, seed bitterness showed no significant relationship with bitterness in wine and a negative relationship to wine dark berry and savory spice flavors. Previous studies showed that wine dark fruit perception can have either a positive relationship (Joscelyne et al. 2007, Bindon et al. 2014) or a negative relationship (Chapman et al. 2005, Lattey et al. 2010, Cadot et al. 2012) with bitterness perception in wine (Bindon et al. 2014). Factors enhancing bitter sensation in wine include alcohol concentration (Vidal et al. 2004, Oberholster et al. 2009), low molecular weight phenols (Casassa and Harbertson 2014, Gonzalo-Diago et al. 2014), and extended maceration (Yokotsuka et al. 2000). The relationships among seed bitterness, wine savory spice, and dark berry sensation found here should be used only as potential indicators of dark berry and savory spice flavors in wine until the diverse factors likely involved in those complex relationships can be determined.
In 2011, several relationships were revealed between pulp detachment from the skin and rim color (r = 0.76, p = 0.00), body color (r = 0.75, p = 0.01), fresh red berry aroma (r = −0.81, p = 0.00), dried fruit aroma (r = 0.66, p = 0.02), alcohol aroma (r = 0.64, p = 0.03), acidity (r = 0.67, p = 0.02), alcohol flavor (r = 0.59, p = 0.05), body (r = 0.61, p = 0.03), tannin quantity (r = 0.78, p = 0.00), tannin quality (r = 0.73, p = 0.006), fresh dark berry flavor (r = 0.72, p = 0.01), dried fruit flavor (r = 0.78, p = 0.00), and savory spice flavor (r = 0.58, p = 0.05). It is possible that some of these relationships are coincidental; therefore, further research is necessary to confirm a causative effect of pulp detachment on some of these sensory attributes. However, a recent study showed that heat stress regimes may significantly affect pulp detachment and that berries grown under heat tents were softer and contained pulp which was easier to detach from the skin (Bonada et al. 2013), while the resulting wines had less anthocyanins, total tannins, and total phenols (Bonada et al. unpublished data 2014). This further demonstrated a positive relationship between difficulty of pulp detachment from the skin and compositional parameters related to wine quality including color, flavor, and mouthfeel.
In contrast to our findings in Shiraz, three previous studies that evaluated Shiraz (Winter et al. 2004), Grenache (Pozzo Di Borgo and Rousseau 2004), and Chardonnay (Rousseau 2001) berries and the corresponding wines at different maturities showed that higher quality in Grenache and Chardonnay wines was achieved when the pulp was easier to detach from the skin. Grenache wines were more full-bodied and had higher prune, confectionary, and pepper aroma character scores, but less red fruit characteristics (Pozzo Di Borgo and Rousseau 2004), while Chardonnay wines were rated higher for white fruits and confit fruit aromas (Rousseau 2001). The correlations found in Shiraz (Winter et al. 2004) highlight that when the pulp is more difficult to detach from the skin, the wines will have more intense vegetal aromas.
The findings of our and the above-mentioned studies agree that pulp detachment from the skin is an important berry sensory attribute to include on BSA score sheets. However, the direction of the relationship of this berry sensory attribute with wine composition and sensory attributes may vary depending on factors such as berry maturity at sample harvest or grape quality.
Relationships between berry sensory attributes, berry composition, and wine composition
In 2011 but not in 2010, pulp detachment from the skin could positively predict wine total tannins (Table 3); a model that is also supported by Pearson’s correlation coefficients (r = 0.82, p = 0.00). After adding berry compositional variables to the PLS regression in 2011, a prediction model was generated where the combination of one berry sensory attribute (pulp detachment from skin) and one berry compositional variable (berry color) could predict one wine compositional variable (wine total tannins) (Figure 2). Both skin detachment and berry color were positive predictors of wine total tannins (Figure 2). The model Q2 coefficient obtained after the inclusion of compositional variables (0.68) was lower than the model Q2 without compositional variables (0.78).
Relationships between berry sensory attributes, berry composition, and wine quality score
PLS regression using berry sensory attributes to predict wine quality scores did not produce a prediction model in 2010, as Q2 and R2 were below the minimum threshold for accepting a model (Table 3). However, PLS regression of 2011 data generated a prediction model where pulp detachment from skin was a positive predictor of wine quality scores (Table 3). Pearson’s correlation also produced a good correlation for pulp detachment from the skin and wine quality scores (r = 0.81, p = 0.00). Previously, Shiraz berries with pulp that was more difficult to detach from the skin were graded higher in the industry allocation of low- and high-grade fruit (Jordans et al. unpublished data 2014).
Although the range of wine quality scores in each study was not large, it was greater in 2011 (14.0 to 15.6) than in 2010 (14.6 to 15.6). The Q2 coefficients generated for the model did not change after including compositional variables. No berry composition variables predicted wine quality scores, when used as a combined predictor with berry sensory attributes in either season.
Relationships between wine sensory attributes and wine composition
Wine quality scores correlated positively in both seasons with wine rim color, wine body color, wine body, wine tannin quantity, wine length, and wine dried fruit flavor (data not shown). Wine quality scores also correlated positively to wine polymeric pigments and wine tannins (Table 4) in agreement with previous findings (Ristic et al. 2010).
Several wine sensory attributes (body color, body, and tannin quantity) also correlated positively with wine polymeric pigments (M3G eq/L) and total tannins (CAT eq/L) in both seasons (Table 4). In Pinot noir wines, a positive correlation was reported between the sensory attributes chemical aroma and earthy aroma and wine pigmented polymers and total tannins (Cortell et al. 2008). Interestingly, wine total tannins correlated positively with fresh dark berry aroma and flavor and negatively with fresh red berry aroma and flavor (Table 4). More recently, positive correlations were found between small polymeric pigments and red and dark wine aroma in treatments to which moderate regulated deficit irrigation was applied (Casassa and Harbertson 2014).
Model reliability: berry sensory attributes and compositional measures as predictors?
When Q2 coefficients were generated from comparing either berry sensory attributes or the combination of berry sensory attributes and composition to three wine parameters (sensory attributes, quality scores, and composition), the Q2 coefficients were very similar. Overall, the models were more reliable when they were generated solely from berry sensory attributes. This suggests that berry sensory attributes alone can reliably predict wine sensory attributes and wine quality scores. The limitation of this study was a low number of samples; however, it is expected that model reliability will increase if a larger number of samples are used and if the range of values for the investigated parameters is wider.
Conclusion
Our study shows for the first time that across two seasons, Shiraz grape berry seed bitterness was a negative indicator of savory spice aroma in wine, an important aroma known to impact wine consumer preference for Shiraz wines. Pulp detachment from the skin showed important relationships with wine quality scores, wine sensory attributes, and wine composition, although statistically significant differences were found in one season only. Given the differences in climate between the two seasons in which samples were collected for this trial, investigation of this attribute over more seasons is required to increase confidence in the possible use of this attribute as a wine quality predictor.
The implied importance of seed bitterness may help grapegrowers and winemakers to identify grape parcels that could potentially produce pepper-like aroma in wines. Furthermore, pulp detachment from the skin can potentially assist with the identification of fruit parcels that may generate wines with more color, dark grape flavor, better tannin structure, and possibly higher quality.
This study showed that there are relationships between berry sensory attributes and some wine sensory attributes that are responsible for wine quality in Shiraz wines. However, more research needs to be conducted using a wider range of grape quality to determine what berry sensory attributes influence wine quality. Research using other grape varieties is necessary to determine whether the relationships are translatable, or if some of these relationships are variety specific. Confirmation of these relationships will not only help wine producers to determine grape flavor potential in the vineyard and thus achievable wine styles; but it could also help producers and researchers to develop vineyard practices and means to manipulate berry sensory characteristics to obtain particular wine styles.
Acknowledgments
The authors thank all the assessors that participated in the grape and wine sensory evaluation. This research was funded by the Australian Grape and Wine Authority. The University of Adelaide is a member of the Wine Innovation Cluster (www.wineinnovationcluster.com) Adelaide, South Australia. The berry and wine sensory tastings were performed with the approval of the University of Adelaide Human Research Ethics Committee.
Footnotes
Supplemental data is freely available with the online version of this article at www.ajevonline.org.
- Received July 2014.
- Revision received November 2014.
- Accepted December 2014.
- Published online May 2015
- ©2015 by the American Society for Enology and Viticulture