Abstract
A trained sensory panel (n = 14) identified key flavors in Sauvignon blanc wines from Australia, France, New Zealand, Spain, South Africa, and the United States. Sixteen characteristics were identified and measured: sweet sweaty passion fruit, capsicum, passion fruit skin/stalk, boxwood/cat urine, grassy, mineral/flinty, citrus, bourbon, apple lolly/candy, tropical, mint, fresh asparagus, canned asparagus, stone fruit, apple and snow pea. Principal component analysis was used to describe differences among regions and countries. Sauvignon blanc wines from Marlborough, New Zealand, were described by tropical and sweet sweaty passion fruit characteristics, while French and South African Sauvignon blanc wines were described as having flinty/mineral and bourbon-like flavors. Chemical analyses of these wines also showed that wines from Marlborough had more methoxypyrazine and thiol compounds. A consumer study (n = 105) showed that New Zealanders significantly prefer New Zealand-style Sauvignon blanc.
The “typicity” for products has been the focus of recent research in Europe (Iaccarino et al. 2006, Martinez Carrasco et al. 2005). The term is used to convey those wine qualities and flavor characteristics that can be expected from a region, which is defined as a broad geographic area distinguished by similar features. In this research, a region is a named area of land. In France, the Appellation d’Origin Controlee (AOC) was established to regulate quality from designated wine-producing regions. Geographic influences on wine sensory profiles have been investigated extensively, including studies on wines made from grape varietals such as Albarino (Vilanova and Vilarino 2006), Touriga Nacional (Falque et al. 2004), Riesling (Douglas et al. 2001), Chardonnay (Schlosser et al. 2005), and Pinot noir (Cliff and Dever 1996). Through the evaluation of sensory characteristics and/or chemical composition, these studies have found regional or subregional differences among the wines. We were interested in determining differences among Sauvignon blanc from different geographical sources in terms of chemical composition and sensory profiles, in combination with consumer preferences. The current study also focused on wines from three regions within New Zealand and compared them with wines produced in five other countries.
Sauvignon blanc wine has distinctive sensory characteristics, both fruity (passion fruit, gooseberry, citrus, tropical) and green (capsicum, asparagus, grassy, leafy) (Cooper 2002). These descriptors have been attributed to key chemical aroma and flavor compounds occurring in the wine. These characteristics have been ascribed to the thiol by-products of yeast fermentation (Tominaga et al. 2000, 1998). The yeast acts upon the odorless thiol precursors in the grapes to produce aromatic thiol compounds in the wine, which have been described as having notes of passion fruit, sweaty, tropical, boxwood, cat’s urine, broom, and grapefruit (Charters 2004, Dubourdieu et al. 2006). It has been postulated that these thiols contribute to the varietal style of Sauvignon blanc wine because they are present in levels much higher than their established perception thresholds (Tominaga et al. 1998). Some yeast produce greater concentration of thiols, making these attributes even more intense (Swiegers et al. 2007). Earlier research has accredited the characteristic capsicum, herbaceous, and green notes in Sauvignon blanc to the methoxypyrazines (Allen and Lacey 1999, Lacey et al. 1991). Unlike the thiols, which only occur after fermentation has begun, the methoxypyrazines are present in the grapes, and their concentrations remain relatively constant during the fermentation process. Methoxypyrazine levels measured in grapes appear to be higher in those cultivated in cooler climates such as New Zealand (Lacey et al. 1991).
One study on closure types and their subsequent effect on the chemical concentrations and flavors of Sauvignon blanc wines demonstrated that, after a year of storage, wines bottled under screwcap underwent very little change in flavor when compared with wines bottled under cork (Brajkovich et al. 2005). The wines with different closures were chemically analyzed for thiols, oxygen, and sulfur dioxide, and then sensorially assessed for six descriptive attributes (capsicum, sweet sweaty passion fruit, passion fruit skin/stalk, cat urine, grassy, flinty/mineral). Another study compared Sauvignon blanc, Chardonnay, and Semillon juice through descriptive analysis (Francis et al. 1994), although the sensory attributes of thiols could not be examined as researchers were evaluating unfermented grape juice and thiols are only present in finished wine. The Sauvignon blanc juice expressed a strong capsicum characteristic in comparison to the other varietal juices.
Winemaker opinions of the Marlborough-style wine were recently evaluated (Parr and et al. 2007). To strengthen the understanding of geographical influences on the flavor characteristics of N.Z. Sauvignon blanc, the current study attempted to provide an objective, scientific correlation of sensorial evaluations with chemical results.
This research of N.Z. Sauvignon blanc began with a narrow assessment of the sensory differences among 28 N.Z. Sauvignon blanc wines selected from the 2003 vintage and found significant differences among the six regions examined (Lund et al. 2005). The study did not include any comparative international samples. Using six sensory attributes to evaluate each wine, the researchers found that Hawke’s Bay Sauvignon blanc wines were high intensity in mineral flinty characteristics, whereas Marlborough wines were high intensity in sweet sweaty passion fruit and capsicum characteristics. Wairarapa wines had higher intensity in cat urine/boxwood characteristics. Some of the wines from specific regions showed measurable differences in their flavor profiles. Based on the results from the N.Z. 2003 vintage, another 35 Sauvignon blanc wines from the N.Z. 2004 vintage were selected from these three regions. Sauvignon blanc wines used in the current study of the 2004 vintage were selected from regions that had shown flavor differences in the 2003 vintage wines (Lund et al. 2005).
Wine marketers and writers claim that Marlborough Sauvignon blanc has distinctive flavors compared with Sauvignon blanc wines produced from other regions (Cooper 2002). In the research presented here, commercially available wines were evaluated to investigate whether Marlborough Sauvignon blanc wine exhibits regionally distinctive flavors as compared with wines from France, Australia, South Africa, Spain, and the United States. Defining the sensory profiles of Sauvignon blanc will aid in understanding the flavors and the chemicals associated with these flavors. Ultimately this research may facilitate the use of chemical measurements to predict descriptive attributes of wine.
Marlborough Sauvignon blanc contributes significant revenue to the N.Z. economy, and the ability to maintain a global position as a market leader for this varietal is critical to the success of the New Zealand wine industry. Scientific exposition of the distinctive flavors of Marlborough Sauvignon blanc may give wine producers the validity to substantiate their marketing claims, and thus benefit the N.Z. economy with increased export sales.
Materials and Methods
Wine
In order to provide a comprehensive sensory evaluation of Sauvignon blanc and to promote a diverse elucidation of definitive flavor profiles, the sensory panel used descriptive analysis to define the sensory characteristics of 52 Sauvignon blanc wines from six countries. Of the 52 wines, 49 were analyzed chemically and eight were selected for further assessment by a consumer panel. The wines were from New Zealand (Hawke’s Bay and Wairarapa in the North Island, and Marlborough in the South Island), France (Sancerre, Loire Valley, Bordeaux), Spain (Rueda), South Africa (Stellenbosch), Australia (South Australia, Western Australia, Victoria), and the United States (Napa Valley, Russian River, and Sonoma in California and Columbia River in Washington). Four to five wines from each country were included in the study, but only two wines could be acquired from Spain (Table 1⇓). Wines were selected on the basis of being predominantly from the Sauvignon blanc grape (>90%). Most of the 52 wines were tank-fermented wine, with little or no oak aging. However, one Hawke’s Bay, one Australian, one American, and two French wines were aged in oak barrels. Oak aging is not a common practice in the production of New Zealand Sauvignon blanc wines, as it is with French Sauvignon blanc. Oak aging reportedly contributes smoky, spice, coconut, and vanilla flavors (Goode 2006). Although oak aging might introduce a confounding effect on the interpretation of the results of this study, several oak-aged samples were included in the descriptive analysis testing, as these wines represent a particular stylistic rendition of Sauvignon blanc available to consumers.
All southern hemisphere wines were selected from the 2004 vintage. The availability of wines from the northern hemisphere at the time of this study was limited to wines from the 2003 vintage, with the exception of one French and two Spanish wines, which were from the 2004 vintage. The retail price of the wines (sometimes used as a proxy for commercial assessment of quality) ranged from US$6.00 to $20.00 per bottle, with most wine prices falling between US$8.00 and $14.00.
Standard chemical wine analysis was performed on all of the wines to attain residual sugar, ethanol, pH, and titratable acidity. Upon completion of the flavor sensory testing, flavor chemical component analyses were conducted on each wine. Wine samples were tested in triplicate for all analyses.
The eight wines chosen for the consumer study comprised a broad range of Sauvignon blanc wine styles, as delineated by the results of the descriptive analysis in the cur rent study. Wines selections were sourced from France, Australia, South Africa, and New Zealand (Hawke’s Bay, Wairarapa, and Marlborough) on the basis of their common commercial availability within the N.Z. market. Marlborough Sauvignon blanc dominates the N.Z. wine market, so three Marlborough wines were included for assessment by the consumer panel. All wines selected for the consumer study were chosen because they represented a distinctive regional flavor profile and not necessarily because they represented what might be regarded as a “typical” regional flavor profile.
Trained panelists
For descriptive analysis, 27 panelists experienced with descriptive analysis were screened for their ability to assess odor and taste, as well as for their cognitive (flavor memory) and descriptive language skills. Fourteen people were selected for the final panel based on their performance for providing correct answers in screening tests. None of the panelists had prior experience in wine assessment. The final panel was comprised of three males and 11 females, ranging in age from 27 to 55 years, and they were paid an hourly wage. Panelists developed the lexicon and reference standards, following normal descriptive analysis (Lawless and Heymann 1999). Panelists completed 70 hours of training in descriptive analysis and in the sensory evaluation of Sauvignon blanc wine.
Consumer panelists
Panelists were recruited on the basis that they were wine consumers. Panelist Sauvignon blanc consumption was evaluated but not used as a selective criteria for recruitment. The authors felt it was more important to understand the preferences and purchasing behaviors of a general wine consumer rather than limit the focus to only Sauvignon blanc wine consumers. Panelists were recruited from wine shops, from the HortResearch workplace, and by word of mouth. Remuneration for participating in the study consisted of a bottle of wine. The 109 consumers evaluated all eight wines chosen for the study.
Facility and evaluation
All sensory testing was performed in booths with green lighting at the HortResearch Sensory and Consumer Science Facility in Mt. Albert, Auckland. A positive airflow was maintained to reduce any odors not associated with the wine. Wine was served at 20°C in standard ISO wine glasses (Gilmours, NZ) with watchglass lids. Double-filtered (Lawless and Heymann 1999) water and plain water crackers were used as palate cleansers. Trained panelists received 20 mL of each wine for testing while consumer panelists received 15 mL of each wine. Both the trained and consumer panel were monadically served samples in a randomized presentation order. The wines were rated on a 150-mm unstructured linescale. The trained panelists rated the intensity of each attribute, from “absent” to “extreme,” on an unstructured linescale. The consumer panel rated their overall liking of the each wine, from “dislike extremely” to “like extremely,” on a 150-mm unstructured linescale. Panelists were permitted to retaste samples if necessary. Consumers were also asked demographic information and purchase behavior questions.
Descriptive analysis
Trained panelists evaluated the 52 wines in triplicate. Panelists evaluated 10 to 11 wines per session, with a 30-sec break after each wine and a 5-min break after every three wines to reduce sensory fatigue. Each panelist returned for 15 sessions so that each individual panelist tasted every wine. Variations were made to the presentation order of wine samples served concurrently to all panelists and to the presentation order of subsequent replicate samples provided to individual panelists.
Assessing 52 wines within a single session cannot be reliably accomplished without encountering the deleterious effects of panelist sensory fatigue. Likewise, when the assessment of a large number of wine samples is scheduled to extend over the course of several panel sessions, there will be the challenge of ensuring that every panelist attend every session. An incomplete randomized block design was applied to manage these challenges. The panelists were given the samples randomly and the randomized samples were blocked by replication (1, 2, 3).
The panelists rated intensities of 16 attributes on computers using Compusense software, version 5.0 (Guelph, Canada). The attributes and their reference standards evaluated are listed in Table 2⇓.
Methoxypyrazine analysis
The quantification of 2-methoxy-3-isobutylpyrazine (MIBP) and 2-methoxy-3-isopropylpyrazine (MIPP) was performed according to a published method (Kotseridis et al. 1999). In brief, the organic phase of a triple extraction of 200 mL of wine (pH 8) with 1:1 diethyl ether:hexane is concentrated down to 100 μL and 2 μL are analyzed by gas chromatography (GC) coupled with mass spectrometry (MS) using a capillary column BP20 (50 m x 220 μm x 0.25 μm). Two modifications were made to this initial method: (1) the use of 2-methoxy-3-([2H3]isobutyl)pyrazine as an internal standard instead of 2-methoxy-3-([2H2]isobutyl)pyrazine and (2) the use of 2-methoxy-3-methylpyrazine as an internal standard for the quantification of MIPP.
The quantification ion of the 2-methoxy-3-([2H3]isobutyl) pyrazine was ion m/z = 127; ions m/z = 154 and 169 were used as qualifiers. For 2-methoxy-3-methylpyrazine, the ion m/z = 124 was used as the quantifier and ion m/z = 106 as the qualifier. The quantification ions of the MIBP and MIPP were ions m/z 124 and 137, respectively, and the ions m/z 151, 164 and 124, 152, respectively, were used as qualifiers.
The standard curve was prepared by adding increasing quantities (from 2 to 50 ng/L ) of MIBP and MIPP to a Sauvignon blanc wine (Marlborough, 2004 vintage) to obtain eight different concentrations. The regression equation obtained was Y = 1077 X − 1.3699 with r2 = 0.9957 for MIBP and Y = 1526.1X + 0.4395 with r2 = 0.9991 for MIPP. Relative standard deviations of 4.8% and 6.2% were obtained for MIBP and MIPP, respectively, by assessing 10 samples of the same wine.
Volatile thiols
An established method (Tominaga et al. 1998, 2006) was used to determine the level of 3-mercaptohexyl acetate (3MHA) and 3-mercaptohexan-1-ol (3MH), using 4-methoxy-2-methyl-2-mercaptobutane as an internal standard. The thiols were extracted from 50 mL of wine using p-hydroxymercuribenzoic acid, which was then fixed onto an anion exchange column before the thiols were eluted with cysteine and extracted into dichloromethane prior to concentration and manual injection of 2 μL onto an Agilent 6890N GC with an 5973 MS detector (Agilent, Santa Clara, CA). The thiols were separated on a 50 m BP20 capillary column (220 × 0.25 μm) using He carrier gas at 28 cm/s and an oven temperature ramping from 40 to 220°C for a 71-min run.
Standard curves were obtained by adding increasing quantities of the two volatile thiols to a Sauvignon blanc wine (50 to 500 ng/L 3MHA; 500 to 5000 ng/L 3MH). The coefficient of determination (r2) was 0.990 for 3MHA and 0.997 for 3MH. The reproducibility of the method was evaluated by repeating the analysis of the same Sauvignon blanc wine six times under constant operating conditions. Relative standard deviations of 6% and 5% were obtained for 3MHA and 3MH, respectively. Thiol extraction was according to a published method (Tominaga and Dubourdieu 2006).
Statistical analysis
Analysis of variance (ANOVA) was determined using residual maximum likelihood (REML), with region selected as the fixed effects and panelist/bottle + region/wine/bottle selected as random effects using GenStat, release 8.1 (Lawes Agricultural Trust, UK). Because of the unequal numbers of wines from each region, standard error of differences (SED) and least significant differences (LSD) vary for each pairwise comparison (Table 3⇓). Principal component analysis (PCA) and canonical variate analysis (CVA) were employed using the fitted wine means for each of the 16 attributes in the descriptive analysis data (SAS Institute, Cary, NC). A one-way ANOVA was used to determine differences between the regional chemical concentration analysis and other standard chemical analysis, such as sugar content and pH, using Fisher’s LSD with 95% confidence level ( p < 0.05).
Partial least squares regression (PLSR) was performed (Unscrambler, version 9.1, CAMO, Oslo, Norway) to determine the relationships among three chemicals and all sensory data. Three of the chemicals (3MHA, 3MH, MIBP) contributed to the prediction of the sensory characteristics, but MIPP did not contribute and was therefore omitted from the PLSR analysis.
The overall liking scores collected from the wine consumers were analyzed using a one-way ANOVA ( p < 0.05) in GenStat. The preference map analysis was conducted in R (R Development Core Team, Vienna, Austria), which took the individual scores of the preference data and projected them into the two-dimensional space of the sensory attributes. A generalized Procrustes analysis (GPA) was performed in R to correlate sensory and consumer data and determine the different clusters of consumers for each flavor profile.
Results
Sensory analysis
Descriptive analysis revealed that the Marlborough wines had distinctive sensory characteristics with intensity levels that exceeded those of the international wines (Table 3⇑). Several attributes (grassy, apple candy, citrus, and canned asparagus) did not show significant p values among different regions. The lack of significance among regions for those attributes was compounded by wide variation in the attribute measurements of wine samples from within a single wine region. Consequently, wines from a specific region may not necessarily display homogenous sensory intensities for those particular attributes.
Principal component analysis (PCA) gives a pictorial relationship of the wines based on their sensory attributes (Figure 1⇓). The PCA simplifies the interpretation of multivariate analyses by extracting two or three dimensions that display the maximum amount of variability among the data. Wines that are very similar appear close to each other. In comparison, canonical variate analysis (CVA) extracts the dimensions that display the maximum amount of variation among the groups of wines from different regions (Heymann and Noble 1989). Results of both the PCA and the CVA were consistent in identifying relevant regional attributes within the data (Figure 2⇓).
With the exception of the wines from Hawke’s Bay, N.Z. regional wines were clearly distinguishable from international wines (Figure 1a⇑). Marlborough and Wairarapa wines showed high attribute intensities for fresh asparagus, sweet sweaty passion fruit, capsicum, passion fruit skin/stalk, tropical, stone fruit, and apple, which comprised most of the variation of the data shown on the x axis (principal component 1; PC1). In contrast, the wines from South Africa, France, Australia, United States, and Hawke’s Bay were characterized by attributes of bourbon, flinty/mineral, and canned asparagus. The variation explained by PC1 was 47.4%. On PC2 (variation explained 14.1%), the wines on the bottom half of the graph displayed more strongly the boxwood/cat urine attribute, while those wines at the top were more intense in the apple lolly/candy characteristics (Figure 1a⇑). To improve the clarity of the plotted data, attributes with joint correlation in PC1 and PC2 of less than 0.5 in absolute value were not labeled on the PCA graph. Although all attributes were included in the analyses, not all attributes are displayed in Figure 1⇑.
Principal component 3 (PC3) (explaining an additional 9.7% variation) further clarified the data (Figure 1b⇑; the attributes on PC1 are the same as in Figure 1a⇑). Wines in the top half of the graph are separated by the presence of asparagus notes (both canned and fresh). Wairarapa wines appeared to have higher levels of both fresh and canned asparagus characteristics; Marlborough wines had more fresh asparagus notes; and international wines had more canned asparagus notes.
The ellipses represent statistical significance at the 95% confidence level around the means of each region (Figure 1a⇑). Because there were only two Spanish wines, they are connected by a single line. The Marlborough mean and ellipse shows no overlap with the international wines, but does show some similarities with the Wairarapa wines.
In CVA, each wine region is represented by a circle, which indicates a 95% confidence interval around the mean score (Figure 2⇑). The Marlborough region produces Sauvignon blanc wines that are significantly different ( p < 0.05) than those from Hawke’s Bay, Wairarapa, South Africa, France, Australia, United States, and Spain. These data suggest that N.Z. 2004 vintage wines had flavor profiles that were distinctive from those of the international wines. The sensory attributes on the left side of the x axis (CVA1) are apple, stone fruit, tropical, passion fruit skin/stalk, fresh asparagus, capsicum, sweet sweaty passion fruit, and cat urine/boxwood, whereas the right side is represented by bourbon and flinty. These are similar attributes to those expressed in PCA1 (Figure 1a, 1b⇑). In PCA (Figure 1a⇑), ellipses of the data from the Wairarapa and Marlborough regions overlap, but that is not the case for the means in CVA (Figure 2⇑). These results occur because the PCA describes the similarities among the individual wines, whereas the CVA assesses differences among the regional means.
Aroma chemical analysis
Chemical analysis was conducted on 50 of the wines in this study (excluding the Spanish wines and one Hawke’s Bay wine, and including a fifth French wine) (Table 4⇓). Marlborough wines were significantly higher in 3MHA (sweet sweaty passion fruit) and 3MH (passion fruit skin/stalk) than wines from all other regions. Wairarapa wines were also high in 3MH and had even higher concentrations of MIBP (capsicum) than wines from other regions. The similarity of asparagus and MIBP green notes may explain the separation of Wairarapa wines, as seen in Figure 2⇑. No differences were found in the concentrations of the MIPP (snow pea) attribute among the wines from the different regions. Although mean concentrations of 3MHA appear high for Marlborough (Table 4⇓), the range in concentration values of 3MHA within the Marlborough wines was also wide, allowing for the possibility that specific wines within the region may indeed have had lower concentrations of 3MHA than wines from other regions.
Relationship between chemical and sensory data
Correlations (r2 > 0.50) for each of three chemical flavor compounds (3MHA, 3MH, MIBP) with their respective sensory attributes are shown in Table 5⇓. The concentration of these thiols can be used to predict the tropical characteristic of wine. The thiols (3MHA and 3MH) had the highest values for the coefficient of determination (tropical, sweet sweaty passion fruit, passion fruit skin/stalk, stone fruit). The tropical reference standard was highly correlated with two chemical compounds 3MHA (r2 = 0.80) and 3MH (r2 = 0.65). The sweet sweaty passion fruit attribute maintained a relatively high correlation (r2 = 0.73) with 3MHA, which was the corresponding sensory reference standard (Table 2⇑). These results support using the chemical measurement of 3MHA to predict the sensory perception of tropical and sweet sweaty passion fruit characteristics. The flavor compound 3MH showed a stronger relationship with the passion fruit skin/stalk attribute (r2 = 0.63), which is the corresponding reference standard (Table 2⇑). Measurement of the concentration of 3MH would predict the sensory perception of passion fruit skin/stalk but not as strongly as using the concentration of 3MHA to predict sweet sweaty passion fruit characteristic in the wine.
The green compound MIBP had the highest positive correlation with the fresh asparagus attribute at r2 = 0.57 and the highest negative correlation with the bourbon attribute (r2 = −0.54). Wines perceived as higher in capsicum, like those from Marlborough, were lower in the bourbon sensory attribute. The reverse was also true, with French wines higher in bourbon and lower in the capsicum sensory attributes. Regional wines that were high in the bourbon characteristics did not necessarily possess high alcohol content. For example, wines from Australia had the lowest mean alcohol at 10.6%, but were still perceived as having relatively high bourbon characteristics. Bourbon was described by panelists as being more of an earthy, smoky character rather than an alcoholic character.
The green compound MIBP had an even higher correlation with the fresh asparagus attribute (r2 = 0.57) than with the capsicum attribute (r2 = 0.37). Although 0.57 is not a high correlation, it does indicate some association with a green character. Wines having higher MIBP concentration will exhibit more fresh asparagus notes. The capsicum character was probably masked by the other components in the wine. Our results (Table 5⇑) confirm other research that described the thiols as passion fruit descriptors (Tominaga et al. 1998, 2000) and that described MIBP as green (Allen and Lacey 1999).
The thiols (3MHA and 3MH) were highly correlated with their associated sensory attributes. These two thiols would serve as better predictors in modeling the sensory profile of wine than MIBP, which has a lower correlation with its sensory attribute, capsicum.
Partial least squares regression highlighted the relationship between the chemical analyses and the trained panel data (Figure 3⇓). The two thiols were shown in close proximity to the sensory attributes tropical, passion fruit skin/stalk, and cat urine/boxwood, which are terms previously used to describe these thiols (Tominaga et al. 1998, 2000, Dubourdieu 2006, Lund et al. 2006). Boxwood has been used to describe high concentrations of 3MHA (Bouchilloux et al. 1998). 3MH may be in close proximity to cat urine/boxwood because 4-mercapto-4-methylpentan-2-one (4MMP) is in the same thiol chemical family. Researchers found that they strengthened their predictive model of Spanish red wines by grouping chemical families on the basis of their sensory and chemical analyses (Aznar et al. 2003). The current study confirms and supports these earlier studies with additional correlations of sensory attributes with chemical composition data.
Wine consumers
Of the 109 consumers, 100% were wine consumers (Table 6⇓). The percentage of women (69%) was higher than the New Zealand percentage of women wine drinkers (55%) (Bruwer 2007). The majority of participants were New Zealanders (69%); other nationalities were Asian, Pacific Islander, European, Sri Lankan, Australian, Indian, and American, none comprising more than 15%. When asked about their white wine preferences and habits, consumers indicated they preferred and regularly drank Sauvignon blanc, followed by Chardonnay. Forty-one percent of the consumers primarily drank white wine, 20% drank predominately red wine, and 39% expressed no preference between red or white wine. When consumers were asked to list the wines they typically drank, 82% noted Sauvignon blanc, 64% noted Chardonnay, and 48% noted Riesling. These consumers (86%) typically spent (US$7.00 to 14.00) (NZ$10.00 to 20.00) on a bottle of wine.
After completing the demographic information and choice questionnaire, the consumers tasted the wines and rated their preference for each wine. The means and ANOVA of their preferences showed these consumers significantly preferred two of the wines from Marlborough compared to wines from Hawke’s Bay, Australia, South Africa, France, and Wairarapa (Table 7⇓). The two Marlborough wines had highest intensities of stone fruit, sweet sweaty passion fruit, cat urine, passion fruit skin/stalk, and tropical, as well as being lowest in bourbon and flinty. The least preferred wine (Wairarapa) possessed average intensities for all the attributes. The French and South African wines were high in mineral/flinty and bourbon characteristics. The Australian wine was highest in apple lolly and lowest in sweet sweaty passion fruit, capsicum, cat urine, passion fruit skin, and fresh asparagus characteristics. The Hawke’s Bay wine was highest in bourbon and mineral/flinty but lowest in tropical, citrus, stone fruit, and apple characteristics.
An external preference map illustrated the sensory space of the wines in relationship to the consumer preference data, and a hierarchal cluster analysis identified groups of consumers and their preferences in relationship to the sensory data (Jaeger et al. 2003). A dendrogram from the cluster analysis identified two distinct groups of consumers (not shown). Cluster 1 indicated a consumer group that preferred a stone fruit, passion fruit skin/stalk, capsicum, sweet sweaty passion fruit, fresh asparagus, boxwood/cat urine-style Sauvignon blanc; whereas cluster 2 consumers preferred a Sauvignon blanc with bourbon as well as flinty/mineral characteristics. Cluster 1 comprised the largest portion of consumers (77%) surveyed and contained a larger percentage (53%) of respondents in the younger age brackets (<34 years) compared with cluster 2. Cluster 1 consumers were more likely to spend over $15 on a bottle of wine (54%) and to be New Zealanders (66%). Divorced people were primarily in cluster 2 and women dominated this cluster (four females to every one male). Eighty-four percent of cluster 1 normally drank Sauvignon blanc as their primary white wine, whereas there were only 68% in cluster 2 who normally drank Sauvignon blanc. Cluster 1 contained a higher percentage of white wine-only drinkers (43%) or those who drank both red and white wines (41%), compared with cluster 2, which had over twice as many red wine-only drinkers (36%).
Discussion
In past research, the Sauvignon blanc flavor profile has been attributed to methoxypyrazines (Allen and Lacey 1999), which give the wine green, capsicum characteristics. However, it has been noted that wines rarely have a sole “impact” compound, such as methoxypyrazine (Noble and Ebeler 2002). Using sensory, chemical, and consumer analyses, the current research determined that the 2004 Marlborough Sauvignon blanc possessed a distinctive and predictable flavor profile that the N.Z. consumers rated as most preferable.
The past literature has enumerated the many attributes associated with Sauvignon blanc wine (Allen and Lacey 1999, Lacey et al. 1991, Dubourdieu et al. 2006, Tominaga et al. 2000, 2006). These attributes (capsicum, grassy, passion fruit skin/stalk, sweet sweaty passion fruit, cat urine/boxwood) are characteristics that were also evident with the wines evaluated in this study. The strongest sensory attributes in Marlborough wines of this study were the high intensities of the fruity and green characteristics, such as tropical, sweet sweaty passion fruit, apple, stone fruit, capsicum, passion fruit skin/stalk, and fresh asparagus. The sensory attributes noted in the wines were highly correlated with the chemical measurements of thiol concentrations. Sensory attributes that contributed less strongly to the Marlborough style were mint, grassy, citrus, and snow pea. The sensory evaluation of snow pea intensities in the wines were confirmed by the chemical measurements of MIPP concentrations. Both analyses showed no significant differences among the wines.
In the sensory portion of this research, the 2004 Marlborough Sauvignon blanc wines not only had green character istics (capsicum, passion fruit skin/stalk, and fresh asparagus), but also high fruity characteristics (tropical, sweet sweaty passion fruit, apple, stone fruit). Statistical analysis of the sensory data (PCA and CVA) demonstrated that the 2004 N.Z. Sauvignon blanc had a distinctive flavor profile which was significantly different from the flavor profiles of the wines from France, Australia, South Africa, United States, and Spain. Although the French, U.S., and South African wines were quite similar, Australian wines were distinguished by their apple lolly/candy characteristic.
The French, South African, Australian, and U.S. Sauvignon blanc wines contained more mineral, flinty, and bourbon sensory characteristics. Analyzing the flavor compounds found in these international flavor profiles, such as 4-mercaptomethyl pentane for the cat urine/boxwood and benzyl methyl thiol for the flinty/mineral overtones (Tominaga et al 1998, 2000) could assist in creating an improved chemically based predictive model.
The chemical concentration of 3MHA and 3MH had higher means in Marlborough wines compared with those from other regions. These high concentrations showed a strong correlation with tropical sensory attributes. 3MHA had high correlation with the sweet sweaty passion fruit, and 3MH was correlated with passion fruit skin/stalk.
Capsicum is a characteristic commonly used to describe Sauvignon blanc, yet within this study MIBP had greater correlation with fresh asparagus than with capsicum. Further investigation might determine what other components could be masking the capsicum attributes in Marlborough Sauvignon blanc.
The sensory data from the 2004 vintage established that Marlborough and Wairarapa wines were somewhat similar, although the latter exhibited stronger asparagus notes. Similar to the results of the 2003 wines, the 2004 vintage from Hawke’s Bay had the lowest concentrations of 3MHA, 3MH, and MIBP compared with the other two regions (Lund et al. 2005). The 2005 vintage has been examined to determine if there is continued consistency among the three vintages.
The Marlborough wines in this study had the highest levels of titratable acidity and residual sugar, the latter only significantly higher than wines from France and Spain. Interestingly, mean titratable acidity levels were significantly higher in all the New Zealand wines compared with the international wines. Increasing acidity is known to diminish perception of fruit characteristics, such as banana, in kiwifruit pulp (Marsh et al. 2006), and when sugar was added, the perception of fruit characteristics increased. Research predicted that an increase in sugar concentration would increase the headspace concentration of “fruity” volatiles in kiwifruit pulp, such as ethyl butanoate and (E)-2-hexanal (Friel et al. 2000). It might be valuable to measure the headspace of Marlborough wines and compare the results to wines with lower levels of titratable acidity and residual sugar.
The chemical data in this research supported the statement that Marlborough Sauvignon blanc wines have a complex style that is not influenced by a single “impact” compound (Noble and Ebeler 2002). There were higher concentrations of thiol (3MHA and 3MH) and methoxypyrazine (MIBP), which created some of the fruity and green characteristics.
The methoxypyrazine of Marlborough Sauvignon blanc has more of a fresh asparagus sensory attribute than a capsicum sensory attribute. Both the 3MHA and the MIBP were more closely associated with a natural product standard (tropical and asparagus, respectively) than with a single chemical as a reference standard (sweet sweaty passion fruit and capsicum, respectively.) The natural product reference standards may more successfully convey a complex sensory perception to a panelist. Perhaps a study evaluating the comparison of sensory reference standards comprised of solely chemical compounds versus reference standards comprised of solely natural products would be of interest in determining whether one set of standards indicates a better prediction of sensory attributes.
The low correlation between MIBP and capsicum character could be explained by a possible masking of MIBP by other components in the wine. Wine is a complex medium, in which many masking and synergistic interactions occur (Peinado et al. 2004). For example, 12% ethanol in water has an extremely strong smell, whereas at the same concentration in wine, the odor is greatly masked by other volatile compounds. Ethanol is capable of masking the perception of esters (Escudero et al. 2007). The negative correlation of the bourbon characteristic to the concentration of MIBP may suggest that there are sensory characteristics that are masked in the presence of compounds such as MIBP. Conversely, the capsicum characteristic may be explained by more than just the chemical concentration of MIBP. A study of sensory and chemical analyses of Spanish red wines found vegetal peppery characteristic to be correlated to isoacids, ethyl esters of isoacids, and fusel alcohol (Aznar et al. 2003). More chemicals will need to be measured and correlated with the sensory attributes to better understand the capsicum perception and the effect MIBP has on the perception of wine aroma.
The thiol and MIBP concentrations could be used to predict a Marlborough style, but it is apparent there are other sensory attributes to consider. Esters such as ethyl decanoate and ethyl hexanoate are also known to be present in Sauvignon blanc wines (Benkwitz et al. 2007). Other flavor compounds, such as esters and C6 compounds, should be measured since they contribute to fruity and green characteristics in wines. Such investigations would enable a more predictive model to be used in anticipating sensory attributes. Studies evaluating synergistic and masking effects of a wider range of chemical compounds would also help in understanding the complex attributes found in wine.
Although there were differences between the wines that could be measured through chemical analyses and sensory evaluation, from a commercial point of view the ultimate consideration is whether the average wine consumer could perceive a difference. Price is less of a dominant predictor of purchasing behavior as wine consumers are becoming more interested in other aspects of wine. Regional reputations are beginning to play a bigger role for the “highly product involved,” more knowledgeable wine consumer (Schamel 2006, Tustin and Lockshin 2001). Consumers in this study preferred wines that presented sweet sweaty passion fruit, capsicum, passion fruit skin/stalk, and fresh asparagus overtones. These results would suggest that N.Z. consumers could recognize and prefer the Marlborough Sauvignon blanc style. One Spanish study found local wines were preferred by locals and purchased on that basis (Martinez-Carrasco et al. 2005). A Spanish consumer study determined that wine origin was more important than price and vintage in influencing consumer selection (Sanchez and Gil 1997). The authors found that while rural consumers desired local wines, urban consumers preferred the perceived higher prestige of wines from the Rioja region, indicating that effects of regionality on consumer behavior are broader than consideration of a wine’s sensory characteristics.
New Zealand wine consumers significantly preferred the unique sensory attributes found in Marlborough Sauvignon blanc wine. These consumers were familiar with Sauvignon blanc, as evident in the cluster analysis results identifying the frequency and selection preferences of their purchasing behavior. The consumers in cluster 1 chose Sauvignon blanc as their most purchased and preferred white wine. In contrast, cluster 2 preferred the flinty, mineral profile of the international wines. Interestingly, cluster 2 had a greater percentage (44%) of non-New Zealanders while cluster 1 had 23%. The research design did not include any determination of how long the non-New Zealander panelists had been residing in New Zealand or the extent of their wine consumption behaviors prior to their arrival. Without this knowledge, only limited conjecture can be made as to whether a limited familiarity with Marlborough Sauvignon blanc may be influencing their wine preference choices. Cluster 2, with more non-New Zealanders, consumed less wine compared to New Zealanders. Sixty-five percent of New Zealanders in the current study consumed wine three or more times per week, whereas only 33% of non-New Zealanders were consuming wine that frequently. Higher wine consumption might infer that these consumers have a greater familiarity with Marlborough Sauvignon blanc and therefore a greater preference, as in the Spanish study (Martinez-Carrasco et al 2005).
According to one study, Australian and New Zealand consumers are increasingly preferring cool-climate wines such as Sauvignon blanc (Schamel and Anderson 2003). Other export markets may not show the same trend in wine preferences. Determining whether international consumers share this cool-climate wine preference will be important to the N.Z. wine export industry. Subsequent investigation of more recent vintages will be important in confirming whether N.Z. Sauvignon blanc is distinct and distinguishable from other regional Sauvignon blanc wines.
Conclusion
Results from sensory analysis, chemical analysis, and New Zealand consumer preference data substantiate the claim that when consumers receive a Marlborough Sauvignon blanc wine, it exhibits distinctive flavors. The 2004 vintage showed significant differences between Marlborough New Zealand and international Sauvignon blanc wines tested in this study. More international wines should be analyzed and tested to confirm these results. Regional differences were also apparent within New Zealand, especially between Hawke’s Bay and Marlborough wines. Wairarapa wines, although similar to those from Marlborough, contained more green characteristics, and consumer data suggested a preference for Marlborough wines. Chemical analysis data showed strong correlations of three chemicals (3MHA, 3MH, MIBP) with some of the sensory attributes. In comparison to methoxypyrazine, the thiols showed higher correlations with the sensory attributes. Investigating the effects of flavor compound masking/synergism may contribute to a more authentic representation of the Sauvignon blanc flavor profile. Lastly, consumers within New Zealand preferred Marlborough Sauvignon blanc to international Sauvignon blanc wines tested in this study. A greater number of international wines should be analyzed and tested to confirm these results.
Footnotes
Acknowledgments: The authors thank the New Zealand winegrowers and wine industry for providing many of the N.Z. wines and Mark Vlossak of St. Innocent winery for procuring some of the international wines. We also thank Lisa Duizer, for her sensory advice, Laurent Jean Francois, who helped in the sensory laboratory, and Denis Dubourdieu and Takatoshi Tominaga for their support and advice in the chemical analysis. The authors also thank the Pernot Ricard team in Blenheim, N.Z., who provided standard chemical analyses on the wines in this study, under the guidance of Andy Frost.
This research was funded by the Foundation of Research, Science and Technology New Zealand (grant UOAX0404).
- Received August 2007.
- Revision received May 2008.
- Revision received July 2008.
- Accepted August 2008.
- Published online March 2009
- Copyright © 2009 by the American Society for Enology and Viticulture