|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 A STUDY OF BLANC DU BOIS WINE QUALITY By ERIC DREYER A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FL ORIDA 2010
2 2010 Eric Dreyer
3 To my parents
4 ACKNOWLEDGMENTS I thank the UF Food Science and Human Nutrition Department and my committee Dr. Goodrich, Dr. Gray, Dr. Welt, and especially Dr. Sims and Dr. Rouseff for their guidance t hroughout this study. I also thank Emma, my family, and my friends for their support. I thank Jack Smoot and June Rouseff at the CREC for their assistance with the chromatography equipment I thank my lab mates Adilia, Lorenzo, Dr. Odabasi, Rene, and Sonia for their help. I thank my wine panelists who stuck with me through 30 sessions of training and tasting. Finally, I thank all the wineries who provided wines for this study.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 7 LIST OF FIGURES .......................................................................................................... 8 ABSTRACT ..................................................................................................................... 9 CHAPTER 1 INTRODUCTION .................................................................................................... 12 2 LITERATURE REVIEW .......................................................................................... 14 Blanc Du Bois Pedigree .......................................................................................... 14 Pierces Disease ..................................................................................................... 14 Blanc Du Bois Growth Characteristics .................................................................... 15 Defining W ine Quality ............................................................................................. 16 Wine Sensory Analysis ........................................................................................... 17 Descriptive Analysis ................................................................................................ 19 Training ............................................................................................................ 20 Wine Studies That Have Employed Descriptive Analysis ................................. 22 Wine Chemistry ...................................................................................................... 24 Alcohols ............................................................................................................ 26 Sugars .............................................................................................................. 28 Volatiles ............................................................................................................ 28 Terpenes .................................................................................................... 30 Hydrocarbons ............................................................................................. 31 Aldehydes .................................................................................................. 31 Ketones ...................................................................................................... 32 Sulfur compounds ...................................................................................... 32 Phenolics ................................................................................................... 33 Amine compounds ..................................................................................... 34 Esters ......................................................................................................... 34 Acids .......................................................................................................... 37 3 METHODS .............................................................................................................. 41 Wine Selection ........................................................................................................ 41 Wine Quality Evaluation .......................................................................................... 41 Descriptive Analysis Panel ...................................................................................... 41 Panelist Selection ............................................................................................. 41 Panelist Training ............................................................................................... 42 Wine Attribute Intensity Evaluation ................................................................... 45
6 Chemical Analysis .................................................................................................. 45 Gas Chromatography Aroma Volatile Analysis ....................................................... 46 Statistical Analysis .................................................................................................. 49 4 RESULTS and DISCUSSION ................................................................................. 52 Quality Judging ....................................................................................................... 52 Descriptive Analysis Ter m Generation .................................................................... 53 Sensory, Chemical, and Volatile Correlations ......................................................... 53 Chemical Analysis .................................................................................................. 64 Principal Component and Cluster Analyses ............................................................ 67 Principal Component Analysis: DA and Chemical Data ................................... 67 Cluster Analysis: DA and Chemical Data ......................................................... 69 Principal Component Analysis: Volatile Data .................................................... 70 Cluster Analysis: Volatile Data ......................................................................... 73 Volatile Content: Similarities to Other Wine Styles ................................................. 73 5 CONCLUSION ........................................................................................................ 95 APPENDIX: VOLATI LE CONCENTRATIONS .............................................................. 97 REFERENCES ............................................................................................................ 101 BIOGRAPHICAL SKETCH .......................................................................................... 107
7 LIST OF TABLES Table page 2 1 Scoring criteria used for the Blanc Du Bois session at the Florida State Fair 21st Annual Wine and Grape Juice Competition ................................................ 39 2 2 Original Davis Scorecard scoring criteria ............................................................ 40 2 3 Updated Davis Scorecard scoring criteria .......................................................... 40 3 1 Final descriptor list and corresponding training references ................................ 50 3 2 Intensity calibration references ........................................................................... 51 4 1 DA attribute intensity, chemical, and qualit y means with Tukeys HSD mean separation1. Wine letter represents quality rank, with A = highest and N = lowest ................................................................................................................. 75 4 2 DA, chemical, and quality correlations significant at p < 0.10 ............................. 77 4 3 DA attribute and volatile correlations significant at p < 0.10 ............................... 79 5 1 Key for identification of volatiles used in PCA on Figure 49 plus Linear Retention Index values for volatiles .................................................................... 94 A 1 Concentrations of volatiles detected by GC MS, in g/L. Odor active volatiles indicated by footnote .......................................................................................... 97
8 LIST OF FIGURES Figure page 2 1 Blanc Du Bois pedigree (Mortensen 1987) ......................................................... 38 2 2 The Dimensions of Wine Quality (Char ters and Pettigrew 2007) ........................ 39 2 3 Monoterpene alcohols and ketones in various wines (Eggers 2005) .................. 40 3 1 Intensity rating scale us ed by DA panel .............................................................. 50 4 1 Quality scores of wine samples as determined by expert judging panel in decreasing order ................................................................................................. 83 4 2 Color meas ured by a spectrophotometer reading absorbance at the 420 nm wavelength. Samples sorted by decreasing quality score .................................. 84 4 3 TA measured in grams of tartaric acid per liter. Samples sorted by decreasing quality score ..................................................................................... 85 4 4 pH of wine samples. Samples sorted by decreasing quality score ..................... 86 4 5 Residual sugar as percent weight of wine samples. Samples sorted by decreasing quality score ..................................................................................... 87 4 6 PCA variables plot showing PC1 and PC2 for the DA attribute intensity data .... 88 4 7 PCA samples plot showing PC1 and PC2 for the DA attribute intensity data. Numbers indicate quality ranking of the wine, with 1 being highest quality ........ 89 4 8 Cluster analysis for the DA attribute intensity data. Numbers indicate quality ranking of the wine, with 1 being highest quality ................................................. 90 4 9 PCA variables plot showing PC1 and PC2 for the GC MS volatile data. Volatiles determined to be odor active using GC O are shaded in. See Table 5 1 for cross reference of volatiles ...................................................................... 91 4 10 PCA samples plot showing PC1 and PC2 for the GC MS volatile data. Numb ers indicate quality ranking of the wine, with 1 being highest quality ........ 92 4 11 Cluster analysis for the GC MS volatile data. Numbers indicate quality ranking of the wine, with 1 being highest quality ................................................. 93
9 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science A STUDY OF BLANC DU BOIS WINE QUAL ITY By Eric Dreyer December 2010 Chair: Charles A. Sims Major: Food Science and Human Nutrition Blanc Du Bois is a hybrid white bunch grape variety developed for its ability to produce high quality white wines, to thrive in the warm, humid climate of the southeastern United States and for its resistance to Pierces Disease. Little is known regarding Blanc Du Bois wine flavor profiles and how these relate to perceived quality. This study investigated the sensory characteristics of Blanc Du Bois wines and used this data to characterize quality differences among the wines. The study was divided into three sections: quality evaluation by expert wine judges, trained panel descriptive analysis (DA), and chemical and volatile analysis of the wines. Eighteen wines from commercial wineries were obtained for the study. All were subjected to judging during a special session at a major wine competition, and quality scores were averaged across the 26 judges ratings. Fourteen of the Blanc Du Bois wines were analyzed b y the DA panel. Fourteen panelists generated a profile of 13 attributes deemed to be the most prominent aromas and flavors in the wines. After training with the aid of references for each attribute and calibrating all panelists with a 15point intensity sc ale, the intensity of each attribute was rated for each wine. Five random wines were presented per session, and each wine was rated in triplicate over
10 the course of the evaluation. The chemical analysis analyzed color, titratable acidity (TA) pH, and resi dual sugar content of the wines. Volatile analyses were performed using static headspace gas chromatography mass spectrometry (GC MS) The DA panel results indicated that the wines were quite var iable in aroma and flavor, with some wines exhibiting charact eristics including tree fruits, citrus fruits, honey, rose, and green character. Wines ranged from very dry to moderately sweet. DA results were analyzed using 2way analysis of variance ( ANOVA) principal component analysis (PCA) cluster analysis, and correlation analysis. There were differences among wines for the intensities of every attribute from the sensory study including aromas such as peach and rose. There were also differences among wines for each chemical assay residual sugar, TA, pH, and col or Correlation analysis indicated that s pecific attributes correlated with high and low quality wines. Wines exhibiting tropical and tree fruit attributes were higher in quality than those with citrus, greenwood/ stemmy and phenolic characteristics. P eac h (0.462) and rose (0.462) correlated positively with quality, while greenwood/stemmy ( 0.678) phenolic/rubber ( 0.555) and bitter ( 0.505) correlated negatively with quality. Of the chemical measurements, only color ( 0.621) had a correlation (negative) with quality at p < 0.10. Correlation analysis also showed that certain sensory attributes correlated with the concentrations of specific volatiles. Fruit attribu te scores correlated primarily with ethyl and acetate esters. Citrus like and green attributes also correlated with certain volatiles, but trends were not clear cut. PCA confirmed that higher quality wines tended to group
11 primarily because of high treefruit and floral sensory scores, while lower quality wines tended to group closer to citrus like green/woody and phenolic scores There is evidence that Blanc Du Bois growing location may influence the aromatic character of the wines, but more work must be done to confirm this apparent trend.
12 CHAPTER 1 INTRODUCTION Blanc Du Bois is a white bunch gr ape that was developed from the grape breeding program at the Central Florida Research and Education Center in 1968. The University of Florida released it for production in 1987 (Mortensen 1987) Blanc Du Bois is n otable for its resistance to Pierce's Disease, early ripening, and the fact that it does not need to be grafted for maximum growth. Blanc Du Bois grapes can produce a very good, spicy white wine given proper production technique (Mortensen 1987) The grape is currently grown in Florida, Georgia, Louisiana, North Carolina, South Carolina, and Texas, where Pierces Disease limits the growth of most other varietals. Mortensen reported in 1987 that Blanc Du Bois was well received at a formal taste panel at Lafayette Vineyards and Winery, grading higher in quality than two of the longer standing Floridagrown white wines, Stover and Suwannee. With a rating of 15.9 out of 20.0, it was placed into the Very Good categor y (Mortensen 1987) In a different sensory evaluation comparing 9 different Florida white bunch grapes, Blanc Du Bois had the highest rating both initially and after aging one year having received scores of approxi m ately 7.0 and 6.5 on a 9point hedonic scale (Sims and Mortensen 1989) Outside of these relatively small and now somewhat dated studies, no formal sensory or chemical data on Blanc Du Bois could be found. It is not known what attributes of Blanc Du Bois wines influence quality, nor is there any information regarding wh ich flavor volatiles influence Blanc Du Bois character. At least 25 wineries are currently maki ng Blanc Du Bois wine, and the majority of them are in Texas. It is not known if flavor and volatile differences exist between Blanc
13 Du Bois wines produced in different regions of the southeast United States. I t has been shown that different soil and climate conditions can cause marked variation in grape g rowth and development, particularly in terms of sugar levels, acidity, and flavor (Reynolds and others 2007, Verzera and others 2008) It would benefit the wine industry to have a better understanding of how different growing conditions influence the character of Blanc Du Bois wine in terms of appearance, flavor, aroma, and chemical composition. Winemaking technique and style also factor into the final character of the wine. For exam ple, consumers may prefer Blanc Du Bois wines finished with a particular residual sweetness level. The objective of this study was to characterize Blanc D u Bois wine sensory attributes in a variety of representative wine samples, evaluate their perceived i ntensities, and identify the flavor and aroma volatiles present in order to determine whether relationships exist between these traits and overall wine quality. The establishment of descriptor terminology for evaluating Blanc Du Bois wines should assist fu ture studies on this wine. It is hoped that grape growers and winemakers will be able to use the information from this study and apply it to their viticultural and winemaking practices in order to improve future Blanc Du Bois vintages. Consequently, produc tion of consistently high quality Blanc Du Bois wines may lead to an increased awareness and recognition of Blanc Du Bois as a desirable white wine. T he wine industry in the southern United States stands to benefit should Blanc Du Bois become more popular, as many wines from this region have thus far remained unknown or been assumed inferior by most consumers.
14 CHAPTER 2 LITERATURE REVIEW Blanc Du Bois Pedigree "Blanc Du Bois" is one of 19 segregants from a cross between Florida D6148, a hybrid that is resi stant to Pierces disease, and Cardinal. D6 148 was a selection from a se lf pollination of Florida A423 (Mortensen 1987) As seen in Figure 21 it is a distant descendent of Vitis aestivalis ssp. smalliana Pixiola (a green grape native to Florida) and Golden Muscat, a Vitis vinifera varietal. The code name for the grape was H1837, but it was later named after Emile Dubois, a 19th century French winemaker who spurred on the Florida wine grape industry by establishing a successful vineyard and winery near Tallahassee, Florida (Woods 2002, Anderson 2006) Pierces Disease Pierce's disease is a bacterial infection of a Xylella fastidiosa strain that uses the glassy winged sharpshooter, or Homalodisca vitripennis formerly known as H. coagulate, as a vector to infect a fruit bearing plant host (University of California Statewide Integrated Pest Management Program (UCIPM) 2008, Mrtensson 2007) Symptoms of Pierce's Disease become evident when the bacteria multiply to such a concentration that they inhibit xylem function in the vine (University of California Statewide Integrated Pest Management Program (UCIPM) 2008) Lethality is high among infected vines, with death occurring 1 to 5 years after infection. Accidentally introduced in the early 1990's, the disease spread rapidly through California and is now found from California to Florida and as far south as Central America (Mrtensson 2007) (Medley 2003) All cultivars of V. vinifera the main wine grape species worldwide, are
15 susceptible, making the disease extremely dangerous to the wine industry (Mr tensson 2007) Blanc Du Bois's resistance to Pierce's Disease is one of the main reasons for its popularity with grape growers and was also one of the reasons the D6148 strain was selected when Blanc Du Bois was being developed. Blanc Du Bois is also res istant to several fungal diseases that often plague vineyards, including downy mildew ( Plasmopara viticola ) and Isariopsis leaf blight ( Isariopsis clavispora ), as well the grape leaf folder moth Desmia funeralis (M rtensson 2007) It is susceptible to other fungal diseases, but preventive fungicide application has been shown to be effective in most cases (Mortensen 1987) Blanc Du Bois Growth Characteristics Blanc Du Bois vin es normally produce about fifty 2.9 gram berries per cluster, yielding an average of 5.3 tons per acre (Mortensen 1987) Data from 2009 estimated total Blanc Du Bois acreage to be approximately 103 acres, with the grapes selling for an average of approximately $900 per ton (Haak 2010) Mortensen assembled data from Blanc Du Bois grapes grown in two separate locations in Florida Leesburg and Tallahassee from 1984 to 1986. T he soluble solids averaged 17.7% with a range of 16.5% to 18.9%, while total acidity ranged from 0.78% to 0.92% with a mean of 0.86%. The pH of the grapes ranged from 3.2 to 3.5, with a mean of 3.35 (Mortensen 1987) Another study in 1986 found a batch of Blanc Du Bois grapes to have 16.9% soluble solids, 1.05% titratable acidity and a pH of 3.56 (Sims and Mortensen 1989) There is no data regarding these parameters for Bl anc Du Bois grapes grown in other states such as Texas, where the bulk of Blanc Du Bois is grown today.
16 Defining Wine Quality Quality is a nebulous term that can take on many different meanings and must be defined in order for data supporting it to have any value (Lawless 1995) Amerine states that wine sensory evaluation can be approached subjectively or objectively, and that ultimately endusers probably lean toward the subjective, or emotional and romantic side, as opposed to the objective, or classical and analytical type of evaluation. This correlates loosely with what Lawless describes as the two types of quality evaluation applicable to foods: quality as consumer appeal versus quality as expert opini on (Lawless 1995) The main indicator of quality as consumer appeal is market performance. Lawless cautions, however, that the product that sells best to the public is not necessarily the same product an expert would select as being highest in sensory satisfaction. Conversely, the main indicator of quality as expert opinion is freedom from defects or deviation from some ideal (Lawless 1995) Charters and Pettigrew explored the dimensions of wine quality through consumers perceptions. Their dimensions of wine quality are shown in Figure 22 demonstrating that wine quality depends on both extrinsic factors related to technical correctness, production, appellation, et cetera and intrinsic the familiar physical attributes that require actually tasting the wine, such as aroma, flavor, balance, and finish dimensions of quality. Furthermore, their findings showed that intrinsic dimensions could be terminal or catalytic. F or example, pleasure and enjoyment gained from consuming the wine was a terminal dimension. The other intrinsic dimensions appearance, aroma and taste, paradigmatic dimensions (a reflection of the wine grapes identity) and the wines potential to improv e with age, were termed catalytic, or
17 indicators which mark out the process of the consumers engagement with the quality of the product (Charters and Pettigrew 2007) This study required that there be a basis on which wine quality is defined. Since the research was concerned not only with quality evaluation but also specific flavor and aroma characteristics and the flavor chemistry influencing those attributes, it made the most sense to define quality in an objec tive manner. Furthermore, assessing Blanc Du Bois wine quality from a subjective point of view would be impractical In the realm of the wine industry, Blanc Du Bois commands such a tiny fraction of market share it would be difficult to find a sizeable contingent of Blanc Du Bois consumers. Wine Sensory Analysis There is much disagreement over what is the ideal method of measuring wine quality. Some consumer minded publications rely on a single evaluator to rate wines on a 50100 point scale (Lawless and Liu 1997) There has been some work with hedonic scaling as well, such as the 14point he donic scale proposed by Lawless and others for generating quality scores for consumer guidance in large scale wine surveys (Lawless and Liu 1997) The 9point hedonic scale, which is immensely popular in food and beverage evaluation, is rarely used in wine studies. For this study, judges at the Florida State Fair 21st Annual Wine and Grape Juice Competition (Tampa, FL, February 2009) used a rating system loosely based on the Davis Scorecard. The criteria and corresponding maximum possible points awarded are listed in Table 21 The predecessor for this rating system was developed in the late 1950s at the University of California at Davis (UC Davis) Department of Viticulture and Enology and was originally designed to be an analytical sensory evaluation method for wine
18 produced from novel grape varietals developed there (Amerine and Roessler 1983) Over the years its use in wine competitions has been popularized, albeit with some changes to the scoring system The original scorecard permitted the rating of the following characteristics by their corresponding possible points awarded, as seen in Table 22 Amidst controversy, the scorecard was l ater modified to the form in Table 23 The ratings for each were: Superior (1720), Standard (1316), Below standard (912), Unacceptable or spoiled (18) Criticism of the use of the Davis scale for this sort of wine judging arose from the fact that a wine could suffer from a serious flaw related to flavor, bitterness, or astringency, yet still receive a Standard grade due to the nature of the scoring sy stem. David Holzgang, the inventor of another scorecard, believes that the Davis scoring system does not place enough emphasis on overall quality and that 2 points is not enough of a spread for general quality. He elaborates that although it is only one aspect of the scorecard, it is important since it is the most subjective category (Holzgang 1981) Another potential pitfall of the Davis scorecard is that a collection of very good wines may yield few if any score differences despite being quite different in character, as long as they all are of sound technical merit (Lawless and Liu 1997) This is due to the fact that most points on the card are based on a point deductionpenalty system for defects. Despite these issues, the Davis scorecard is still used in many wine competitions today, including the prestigious Florida State Fair Wine and Grape Competition.
19 A 14 year study determined that it takes years (approximately 5 in m ost cases) for an expert judge to consistently provide a normal distribution of scores on the Davis Scorecard (Ough and Winton 1976) This is due primarily to the unbalanced nature of the card; there is more room for judging on one side (negative) than the other. This introduces the possibility that some judges may consciously or unconsciously utilize this lower range to a greater extent than others, skewing the relative distribution of scores (Ough and Winton 1976) Descriptive Analysis D escriptive analysis (DA) is a sensory evaluation method used extensively in the food and beverage industry for acquiring qualitative and quantitative data regarding product taste, aroma, texture, and appearance (Lawless and Heymann 1998, Meilgaard, Civille and Carr 2007) There are several end uses for the data in the areas of product development, quality control, shelf life, and competitor product comparison. The ability of DA to identify and quantify product attributes and permit correlation w ith instrumental analysis and quality data was particularly appropriate for this study. DA differs from and complements instrumental analysis through the type of data that is collected. Perceived f lavors and aromas are often produced by more than one chemical compound, and since odors are not additive variables, one cannot predict the identity of a sample aroma by simply identifying chromatographic peaks on a gas chromatography olfactory (GC O) unit (Carlucci and Monteleone 2008) A DA panel, on the other hand, detects specific combinations of volatile compounds as learned and identifiable aromas. The number of judges on a DA panel is usually dependent on the nature of the product being studied. Simple products may have panels as small as 5 subjects,
20 whereas more complex products or samples with smaller differences between treatments may req uire larger subject numbers (Meilgaard, Civille and Carr 2007) As described in Chapter 3: Methods the DA panel in this study used a hybrid of two popular DA methods: Quant itative Descriptive Analysis (QDA) and The SpectrumTM. There are distinct differences in the protocols for each method notably, panel leader involvement, intensity scaling/scoring, and attribute/terminology development (Meilgaard, Civille and Carr 2007) In QDA, the panel leader does not exert much influence on the panelists other than to ensure that they using the same terminology. This leaves the panelists to interpret the intensity rating and scale usage themselves, as long as they are consistent (Meilgaard, Civille and Carr 2007) The scale itself is a 15 cm line scale. In the Spectrum method, the panel leader is extensively involved in training the panelists to become explicitly familiar with Lexicons, or arrays of standard attribute names (Meilgaard, Civille and Carr 2007) These attribute names are usually selected prior to the start of the p anel; the vocabulary is not always generated by the panelists (Lawless 1995) Training Panelist training can span from just one week for judges with extensive DA and wine tasting experience to 9+ weeks for panels w ith variable experience levels. In most wine DA panels, the actual amount of time spent per week was several hours (Carlucci and Monteleone 2008, Mirarefi, Menke and Lee 2004, Blackman and Saliba 2009) The Wine Aroma Wheel serves to assist panelists with term generation and communication. It was developed at UC Davis in 1984 (Noble and others 1984) The wheel contains three tiers of descriptive terms, with the 12 most general in the center tier, which resembles a pie chart. The second tier contains 29 terms used to split a
21 general term, such as "fruity," into more specific descriptors, such as "citrus," berry, tree fruit, "tropi cal fruit," dried fruit, and other (Noble 1987) The third tier contains actual aroma descriptors such as "pineapple," "melon," and "banana" for the category "tropical fruit." If the panel is performing term gener ation, this method is advantageous since it eliminates quality or likingbased descriptors and encourages the panelists to use more objective language in the term generation stage (Noble and others 1984) The wine wheel has been successfully employed in wine tasting courses as well as DA studies (Carlucci and Monteleone 2008, Mirarefi, Menke and Lee 2004) The wine wheels benefits are enhanced by the use of appropriate reference standards (Noble and others 1987) Panelists involved in a QDA type panel work with the panel leader and each other to create a set of standards that are suitable references for both training and intensity evaluation (Lawless and Heymann 1998) Another tool used in training DA panelists for wine evaluation is an aroma reference kit. The kit used in this study Le Nez Du Vin contains 54 references r epresenting the most common aromas found in wine (Lenoir 2006) Specialized kits are also available, such as a 12 sample kit cont aining references found in oak barrel aged wines and another 12 sample kit with references representing common wine faults (Lenoir 2006) The references in Le Nez Du Vin which exist as either natural essences or synthetic mixtures of compounds, are designed to be stable over time and are contained inside small, screw cap glass vials (Lenoir 2006) Their actual stability has been brought into question by at least one study (Noble and others 1987) Past wine evaluation studies have relied heavily on these kits for training and as reference standards (Sauvageot and Vivier 1997)
22 Wine Studies That Have Employed Descriptive Analysis A wine study that employed a simil arly designed hybrid DA methodology was perf ormed by Mirarefi and others (Mirarefi, Menke and Lee 2004) The authors tested 12 wines made from a hybrid grape called Chardonel. Thei r research aimed to develop a lexic on and standard references for Chardonel wines and to characterize Chardonel wines from different states in the Midwest  by the descriptive terms developed (Mirarefi, Menke and Lee 2004) The study utilized 13 judges with no previous formal wine evaluation experience, and 24 training sessions were held, split into term generation and intensity rating sessions. Panelists were given a reasonable amount of control over the term generation. Based on consensus agreement, attributes that were very difficult to detect or that were present at equal intensities across all samples were excluded. References were made available for all 23 terms in the study (Mirarefi, Menke and Lee 2004) ANOVA was used to determine whether wines were a source of variation among the 23 attributes. This is a widely practiced statistical method among DA panels, as it provides the most obvious feedback on whether differences existed between the wines, or whether variation in their scores was due to random or panelist error (Carlucci and The authors then performed correlation analysis to see which attributes were positively or negatively correlated. Cluster analysis was also performed on the wines that differed significantly in order to see if any t rends existed between the cluster a nalysis and PCA results (Mirarefi, Menke and Lee 2004) The most obvious correlation in their results indicated that increased oak barrel aging time w as positively correlated with inc reased intensity of the oak aroma attribute.
23 In the PCA analysis, PC1 (62.4%) contrasted wines high in Granny Smith apple flavor, grapefruit aftertaste, sour flavor, bitter flavor, and astringent texture/mouthfeel attributes with those wines high in sweet pear, and Jonagold apple flavor attributes (Mirarefi, Menke and Lee 2004) According to the PCA plot, the location where the Chardonel grapes were grown did not appear to influence the grouping of the wines (Mirarefi, Menke and Lee 2004) There are numerous other studies that have used DA A study by Lund and others (2009) characterized Sauvignon Blanc wines from 6 different countries and explored relationships between the sensory properties, chemical data, and trends from a consumer study that determined the demographic information of New Zealand wine consumers (Lund and others 2009) The study concluded that there are two distinct groupi ngs of Sauvignon Blanc wines, depending on their origins: those with tropical and sweet sweaty passion fruit characteristics and those with flinty/mineral and bourbonlike flavors (Lund and others 2009) A stu dy by Skinkis and others (2010) used DA to characterize the flavor and aroma differences among wines made from two vintages of Traminette grapes that were grown under variable sunlight levels. Panel ist s rated the wines made from grapes with the highest sunlight exposure as having increased aromatic intensity for several attributes. These results supported their chromatographically determined findings that the exposed grapes contained higher concentrations of potentially volatile monoterpenes (Skinkis, Bordelon and Butz 2010) Another study, by Blackman and Saliba (2009) used DA to characterize Hunter Valley Semillon wines. Aging of this Australian style is common practice, and the
24 researchers were interested to learn if and how the maturation process affected the wine character (Blackman and Saliba 2009) Their panel of 15 trained judges identified and rated the intensities of 12 aromas as well as acidity and sweet ness f or 16 wine samples. The samples vintages spanned from 1996 to 2006. The authors found that their PCA divided the wines into four distinct groupings, with bottle age driving the separation. The researchers had knowledge of some of the viticultural and winemaking practices employed for the wines in the study and thus were able to forecast that one of the four groups would likely transition to a different group after several years of cellaring (Blackman and Saliba 2009) Wine Chemistry Wine is thought to be the secondmost complex known liquid next to human blood, due in part to the myriad biochemical reactions that occur during production and aging and in part to the sheer number of volatile compounds present which reaches into the hundreds (Gurban and others 2006) Wine in its most basic terms is comprised of water, ethanol, glycerol, proteins, polysaccharides, aroma compounds and volatiles (Jones and others 2008) Ethanol creates a sensation of fullness, hotness or burn in the mouth and contributes to viscosity (Amerine and Roessler 1983, Pick ering and others 1998) while also serving as a solvent for many aroma compounds (Lenoir 2006) Glycerol, the most abundant nonethanol byproduct of wine strains of Saccharomyces cerevisiae is usually considered to be a viscous material, but at its relatively low concentration of about 7 g/L in wine it may enhance the perception of sweetness (Amerine and Roessler 1983, Ya
25 2007) Several other studies have provided evidence that it may also increase viscosity to some degree (Jones and others 2008, Nurgel and Pickering 2005) Proteins from the yeast and grapes are normally present in concentrations far below the organoleptic detection threshold and are not considered to contribute to wine flavor or aroma (Jones and others 2008) Polysaccharides have been shown to contribute to the mouthfeel of wines and, according Pellerin and Cabanis (as cited in Vidal and others 2004), are comprised mostly of type II arabinogalactanproteins, rhamnogalacturonans and yeast derived mannoproteins (Vidal and others 2004, Vidal and others 2003) A study by Jones and others (2008 ) examined the sensory properties of a synthetic wine system created from t he aforementioned collection of components. An aroma compound stock solution was prepared using 14 volatiles that were identified by GC MS as being present in levels exceeding their sensory detection thresholds. These compounds were added to the stock solution in the approximate concentrations found in the original wine sample. According to the sensory analysis, the addition of these volatiles produced significant effects in overall aroma, estery aroma, floral aroma, overall flavor and acidity at p < 0.05, as well as hotness at p < 0.10. The researchers tested those attributes in addition to cheesy aroma, citrus aroma, peach aroma, overall flavor, sweet, bitter, viscosity, metallic, drying, and texture attributes with many combinations and permutations of polysaccharides, proteins, glycerol, ethanol, and volatiles being present in the synthetic wine system samples ( Jones and others 2008) Their results indicated that there were very few instances where particular co mponents were implicated in main effects that did not involve higher order
26 interactions ( Jones and others 2008) Thus, the wine flavor was not dictated by the odor active volatiles they had added. For example, they found that ethanol, glycerol, protein, and polysaccharides all had significant effects on orthonasal perception at p < 0.10 and in some cases at p < 0.05 With that in mind, the authors reported that it was the volatiles including a number of esters, o damascenone and ethanol that had the greatest impact on most of the aroma and textural attributes for the model system ( Jones and others 2008) Alcohols Alcohols constitute a large fraction of volatile wine aroma compounds, ranging from the simplest and smallest (methanol) up to much larger monoterpene alcohols and beyond. These compounds impart a wide range of aromas to wine, from grassy to fruity to meaty to spicy (Rouseff and Smoot 2006) Ethanol is the most prominent volatile in wine, ranging from 521% by volume depending on wine style. It is a byproduct of yeast metabolism of sugar during alcoholic fermentation (Amerine and Roessler 1983) Most table wines, including the samples e xamined in this study, contain between 11 and 13% ethanol by volume (Amerine and Roessler 1983) As explained ear lier in this section, ethanol plays an important role in mouthfeel, flavor, and aroma of wine (Jones and others 2008) It has a relatively high detection threshold; one study found it to be approximately 17 g/L (Meilgaard 1993) Its flavor by mouth is perceived as somewhat sweet (Amerine and Roessler 1983) Far lower concentrations of methanol are also present to some degree in wine, but outside of extreme cases this alcohol does not have a direct sensory impact (Amerine and Roessler 1983) The same is the case with glycerol, a triol present in relatively large
27 quantities (0.22.0% (Amerine and Roessler 1983) ), whose sensory effects, as previously discussed, are both minor and disputed. There are numerous other alcohols in wine. Many of them are small, aliphatic compounds and may be char acterized by more or less of the fusel oil odor which, when present at high concentrations, impart an unpleasant character to wine. (Amerine and Roessler 1983) Others, such as 1and 3hexanol, may contr ibute grassy, floral, or winey aromas (Rouseff and Smoot 2006, Torrens and others 2010) The concentrations of these alcohols are highly dependent on both the initial content within the grapes a nd the winemaking process; low fermentation temperatures suppress their formation, while the opposite is true for higher temperatures (Amerine and Roessler 1983) Other prominent alcohols found in grapes and w ine include phenethyl alcohol, which gives a floral or roselike aroma (Lenoir 2006) and isoamyl alcohol (3methyl 1 butanol), which gives a malty or burnt aroma (Amerine and Roessler 1983, Rouseff and Smoot 2006, Acree and Arn 2004) 2,3butanediol is another common alcohol, but its contribution to wine is disputed due to its high sensory threshold (Amerine and Roessler 1983, Bartowsky and Henschke 2004) 3 methyl 1 pentanol is found in many types of wine (Zea and others 2001, Komes, Ulrich and Lovr ic 2006) and contributes a green or winelike aroma (The Good Scents Company 2010) Sugar alcohols such as sorbitol, mannitol, erythritol, and arabitol are found in wines at levels usually not exceeding 400 mg/L for all except mannitol, which is normally the most prominent (Amerine and Roessler 1983, Amerine, Ough and Ough 1980)
28 Sugars Glucose and fructose in wine may be present as unfermented, residual sugars or as a sweetener added to wines. Humans are more sensitive to and thus perceive fructose as being sweeter at a given concentration, although this perception is not linear and changes with increasing fructose concentration (Amerine and Roessler 1983, Damodaran, Parkin and Fennema 2007) Different grape varietals and grapes grown and harvested under different conditions may exhibit varying glucose/fructose ratios. Increased sugar concentrations tend to decrease the intensity of perceived sourness in wine, al though this phenomenon differs among individuals (Amerine and Roessler 1983) In dry wines few of these sugar molecules are left unfermented, so their flavor contribution is limited outside of the indirect effect that the glucose/fructose ratio may have on the byproducts produced by yeast metabolism. Volatiles The complexity of wine volatile profiles makes them notoriously difficult to analyze (Barbe, Pineau and Ferreira. Antonio Cesar Silva 2008) It has been shown, however, that most wines are comprised of a fundamental set of volatiles that occur in high conc entrations, along with a larger and more diverse array of compounds in lower concentrations (Amerine and Roessler 1983) This does not necessarily mean that the compounds with lower concentrations have less im pact on a wines character (Lenoir 2006) Avakyants and others reported in 1981 (as cited in Amerine 1983) that, "The basic odor of wines is attributed to four esters (ethyl acetate, isoamyl acetate, ethyl caproate, and ethyl caprylate); two alcohols (isobutyl and isoamyl); and one aldehyde (acetaldehyde). There are no more recent studies that confirm this using a modern technique such as GC O.
29 Due to the nature of human olfactory perception, there is a threshold eff ect, which states that an individual cannot detect a volatile compound unless it is present above its threshold value in units of concentration (Lawless and Heymann 1998) That value can vary from person to pers on, so sensory thresholds are studied with this possible variation in mind. What one judge might detect during one tasting might never be detected by another judge (Lawless and Heymann 1998) Wine aroma compounds are commonly divided into their origin relative to the step in the winemaking process during which they formed. The first category is primary aromas, also known as varietal or grape aromas. These compounds are found in the fresh, uncrushed grapes, and their profiles vary widely from varietal to varietal (Lenoir 2006, Rapp 1990) The second category is the secondary aromas. Lenoir combines both prefermentation and fermentation aromas as secondary aromas, while Rapp considers them to be only those formed through chemical, enzymatic, and thermal reactions during maceration, pressing, and other must production processes (Lenoir 2006, Rapp 1990) Fermentation aromas can be formed as byproducts of yeast metabolism during alcoholic fermentation and also through bacterial metabolism during malolactic fermentation, particularly in the form of acids, esters, aldehydes, ketones and [sulfur] compounds (Rapp 1998) Additionally, glycosides can be a source of odor compounds in wine. These compounds are found often in grapes and usually consist of an aroma volatile bound by a carbohydrate that prevents the volatile from having aroma activity (Noble and others 1988) During fermentation and aging, acid or enzyme catalyzed hydrolysis of the glycoside frees the volatile ( Sefton, Francis and Williams 1993,
30 Reineccius 2006) Reineccius wrote that this process may be the key to accelerating the aging of fine wines, though attempts to do so artificially by means of enzyme addition, heating, or acidification have all failed. T he final wine aroma category is the products of various chemical reactions that occur during barrel and/or bottle aging. These could lead to decreased or increased concentrations of certain aroma compounds present in the wine (Lenoir 2006, Rapp 1990) or added aroma compounds in the case of barrel aging. Terpenes Terpenes are a class of hydrocarbons that are characterized by their five carbon isoprene unit base structures (Reineccius 2006) Many variations of the structure exist, including the number of isoprene units, degree of unsaturation, ring formations, and oxygen, nitrogen, and sulfur content. These variations give rise to the varied sensory effects te rpenes exhibit. Monoterpenes, which consist of 3 isoprene units (15 carbons) may be present in wines in concentrations up to 6 mg/L (Mateo and Jimnez 2000) Their odor impact is essential to the characteristic aroma of muscat grapes and wines, and varies for other grape varieties (Amerine and Roessler 1983) Terpenes are often present in concentrations below their sensory threshold, rendering them undetectable by the human nose, as seen in Figure 23 High levels of linalool, a terpene alcohol, are characteristic of muscat wines (Mateo and Jimnez 2000) and are known to impart a fruity, floral character to wine (Lee and Noble 2003) The most frequently encountered terpenes in wine, besides linalool, are geraniol, nerol, and linalool oxides (Amerine and Roessler 1983, Mateo and Jimnez 2000) all of which exhibit a floral aroma (Rouseff and Smoot 2006) although other mo terpineol, hotrienol, citronellol, nerol oxide, myrcenol,
31 and ocimenol are not uncommon (Amerine and Roessler 1983, Mateo and Jimnez 2000) Terpenes are relatively stable throughout fermentation (Amerine and Roessler 1983) but degrade slowly during bottle aging (Rapp 1998) which subsequently will affect the sensory properties of a wine as it matures Terpene alcohols represent some of the volatiles commonly bound up as glycosides that exist as nearly tasteless compounds in wine until they are hydrolyzed and released (Noble and others 1988) Hydr ocarbons Vitispirane is the common name for 2,6,6 trimethyl 10methylidene1 oxaspiro[4.5]dec 8 ene, a C13 norisoprenoid compound formed from carotenoid degradation (Eggers 2005, Rapp 1998) There are two stereoisomers of the compound, each exhibiting distinctive and different floral aromas (Eggers 2005) Multiple studies have confirmed that the concentration of this compound increases as wine ages (Eggers 2005, Amerine and Roessler 1983, Torrens and others 2010) Aldehydes Few aldehydes exist in wine because their carbonyl group is reactive and prone to reduction duri ng fermentation (Verzera and others 2008, Amerine and Roessler 1983, Reineccius 2006) Acetaldehyde is a common product in wine fermentations, but primaril y exhibits its oxidized wine note when present at higher concentrations, such as in sherry (Amerine and Roessler 1983) Decanal is a prominent compound in musts, but has also been found at lower levels in wine and is described as grassy and arugulalike (Torrens and others 2010) A number of aldehydes appear in wine due to oak barrel exposure (Lee and Noble 2003) Hexanal and tra ns 2 nonenal are present in the green wood standard of Le Nez
32 Du Vin New Oak aroma reference kit. Hexanal is a byproduct formed from the oxidation of lipids and has a pungent fatty, green character. The accompanying literature claims that the trans 2 nonenal aroma can be interpreted as cucumber (Lenoir 2006) According to Lenoir, vanillin, an aldehyde, and syringaldehyde are both products derived from the woodaging process, and both exhibit vanilla aroma (Lenoir 2006) Furfural, or furaldehyde, in low to moderate quantities is said to contribute smoothness to wine and some burnt sugar aroma. It is a breakdown product of xylose (Lenoir 2006) which explains its higher levels in barrel aged wines. Ketones A number of ketones are prominent in sherry wines and ports, as reported by Schreier (1979) and Simpson (1980), respectively, as cited in Amerine (1983). Amerine explains that some ketones that are found in table wines are 3hydroxy 2 butanone, 2,3pentanedione, and 3hydroxy 2 pentanone, but that they seem to have little sensory impact (Amerine and Roessler 1983) Diacetyl, or 2,3 butanedione, which is most frequently found in red wines and exhibits buttery (Rouseff and Smoot 2006) and oxidative notes, is another notable ketone (Amerine and Roessler 1983) i onone and demascenone have been found in red and white wines and give distinct raspberry and floral aromas (Lenoir 2006, Acree and Arn 2004, Kotseridis and Baumes 2000) Sulfur compounds Sulfur compounds may be a result of natural production or winemaker error. Excess sulfur dioxide added during the vinification process can get reduced to hydrogen sulfide, a notoriously foul smelling agent that is considered a wine defect (Lenoir 2006) Dimethyl sulfide, which smells of sulfur, cabbage, and mold (Rouseff and Smoot 2006,
33 Acree and Arn 2004) is produced by yeast metabolism from cysteine, cystine, and glutathione (Amerine and Roessler 1983) Odor active sulfur compounds tend to have very low sensory thresholds, with dimethyl sulfide becoming detectable at 0.31 .0 (Leffingwell and Associates 1999) Other sulfur compounds may exhibit oniony or fruity aromas (Lenoir 2006) 1 p menthene 8 thiol, (Rouseff and Smoot 2006) 4 mercapto 4 methanol 2 pentanone is another potent, fruity smelling sulfur volatile found primarily in white wines. Phenolics A vast array of phenolic compounds exist s in both red and white wines and may be present in the grapes or produced by yeast or bacteria during fermentation (Amerine and Roessler 1983, Lenoir 2006) Most exhibit similar odors that have been described as phenolic, plastic, medicinal, and musty (Acree and Arn 2004) al though others can be reminiscent of smoke or leather (Lenoir 2006) The aroma thresholds for phenolic whereas 4 vinylguaiacol is (Leffingwell and Associates 1999) Lenoir explains that the concentration in wine is crucial, giving the example that approximately 2 mg/L of 4ethyl phenol in wine gives the elegant scent of leather, but at 4 mg/L it smells of horse manure (Lenoir 2006) Polyphenolic compounds in wine, such as tannins in red wine (Fontoin and others 2008) are a source of bitterness, a taste sensation, and astringency, a tactile sensation (Amerine and Roessler 1983) These compounds tend to exhibit a binding effect with salivary proteins, creating a highly hydrophobic layer that causes the customary mouthdrying phenomenon (Fontoin and others 2008) The study by Fontoin indicated that the
34 polyphenolic concentration did not necessarily dictate perceived astringency, because ethanol concentration and pH of the wine also had significant effects (Fontoin and others 2008) Amine c ompounds Many amines and N acetylamines have been reported in wine, but their significance to wine flavor and aroma is negligible in most cases (Amerine and Roessler 1983) There is evidence of biogenic amine production in white wines, but the concentrations tend to be higher in red wines (Herbet 2005). Esters While certain esters have no odor activity (Amerine and R oessler 1983) it is thought that as a functional group they are more important to the flavor of alcoholic beverages than any other class of compounds (Reineccius 2006) Esters are generally produced as indirect byproducts of fermentation, formed by the esterification of acids and alcohols by yeast. Given that most esters and ethyl esters are formed from these precursors, it is not surprising that the esters with the highest concentrations are formed from the aci ds and alcohols that were in the highest concentrations in grapes: ethanol, isobutanol, and isopentanol (Amerine and Roessler 1983) Vianna and Ebeler described the process concisely, explaining that, Fatty acid ethyl esters (e.g., ethylbutanoate, ethylhexanoate, ethyloctanoate, etc.) are obtained from ethanolysis of acylCoA that is formed during fatty acid synthesis or degradation. Acetate esters (e.g., isoamyl acetate, propyl acetate, hexyl acetate, phenethy l acetate) are the result of the reaction of acetylCoA with higher alcohols that are formed from the degradation of amino acids or carbohydrates (Vianna and Ebeler 2001)
35 Common esters found in wine gas chromat ography studies include ethyl butanoate, ethyl 2methyl butanoate, ethyl 3 methyl butanoate, ethyl pentanoate, ethyl hexanoate, ethyl heptanoate, methyl and ethyl octanoate, ethyl nonanoate, methyl and ethyl decanoate, and ethyl dodecanoate (Verzera and others 2008, Lee and N oble 2003, Fan and others 2010) The ethyl esters generally have very low sensory h ave higher thresholds of approximately 20(Leffingwell and Associates 1999) The smaller chain methyl and ethyl esters (up to 5 carbons) such as ethyl butanoate and ethyl pentanoate tend to be characterized as fruity and/or applelike. Some of the midsized compounds exhibit fruit characters such as banana, peach, apricot, apple, or even winelike aromas. The larger esters, such as ethyl decanoate and dodecanoate, may exhibit fruity, waxy, or fa tty odors and tend to have higher sensory thresholds (Rouseff and Smoot 2006, Acree and Arn 2004, The Good Scents Company 2010, Leffingwell and Associates 1999, Vilanova and Sieiro 2006) Other esters are also found in wine, including diethyl succinate (wine, fruit), ethyl lactate (fruity), and acetate esters such as isoamyl acetate (banana) and hexyl acetat e (fruit, herb) (Verzera and others 2008, Rouseff and Smoot 2006, Lee and Noble 2003, Fan and others 2010) The first two esters have relatively high thresholds, in the 10000 (Leffingwell and Associates 1999, Perestrelo and others 2006) Ethyl lactate (fruity, butterscotch) and diethyl succinate (fruity, winey) (Acree and Arn 2004, The Good Scents Company 2010) usually increase in concentration during oxidative aging (Zea and others 2001, Prez Prieto, Lpez Roca and Gmez Plaza 2003)
36 A study by van der Merwe and van Wyk involved adding a combination of 6 purified esters to deodorized w hite wine and analyzing the quality and intensity of the wines odors (van der Merwe and van Wyk 1981) The esters selected (isoamyl acetate, n hexyl acetate, 2phenethyl acetate, ethyl n hexanoate, ethyl n oct anoate, ethyl n decanoate) were prominent in the Chenin Blanc model wine and were added back in concentrations relative to the model wine concentration. These synthetic wines were evaluated with only ethyl esters added, only acetate esters added, both added, and as full sets minus individual esters. The more complex ester additions yielded more significant differences in quality and intensity than the less complex additions, and the singleester removed trials yielded no differences among the 6 treatments (van der Merwe and van Wyk 1981) Several subexperiments were performed on the compounds present in highest concentrations: isoamyl alcohol, isobutanol, and ethyl acetate. Ethyl acetate caused a negative effect on quality when present at high levels, while neither isoamyl alcohol nor isobutyl alcohol had any quality or intensity effects on the wine aroma when added at levels above those normally present in the wines (van der Merwe and van Wyk 1981) Ethyl acetate is known to exhibit both solvent and overripe fruit aromas (The Good Scents Company 2010) Various lactones have decalactone. Lactones contribute to a number of different aromas, including sherry, whiskey (Amerine and Roessler 1983) peach (Lenoir 2006) and coconut (Fan and others 2010)
37 Acids Organic acids generally play a minor role in wine aroma due to their relatively low volatility (Amerine and Roessler 1983) Some acids tartaric and malic in particular exist in the grapes, whereas others are byproduct s of yeast metabolism and are negligible in must (Hutkins 2006) Tartaric and malic acid are extremely important to wine flavor, as wine pH and perceived acidity are directly related to their concentrations. Wines with too low or high pH are considered inferior (Hutkins 2006) Furthe rmore those with a high pH might be more susceptible to microbial contamination, which could lead to further off flavors. Succinic acid and lactic acid may also contribute to acidity. The most prominent aliphatic acids in wine include formic, acetic, octanoic, and decanoic acid although of these only acetic acid is normally odor active (Amerine and Roessler 1983)
38 Figure 21. Blanc Du Bois pedigree (Mortensen 1987)
39 Figure 2 2 The Dimensions of Wine Quality (Charters and Pettigrew 2007) Table 21. Scoring criteria used for the Blanc Du Bois session at the Florida State Fair 21st Annual Wine and Grape Juice Competition Cri terion Score (points) Color/Clarity 2 Aroma 5 Flavor 4 Balance 5 Overall Quality 4 Comments Medal Double gold, Gold, Silver, Bronze, No medal
40 Table 22. Original Davis Scorecard scoring criteria Criterion Score (points) Appearance 2 Color 2 Aroma and bouquet 4 Total Acidity 2 Volatile Acidity 2 Sweetness 1 Body 1 Flavor 2 Bitterness 2 General quality 2 Table 23. Updated Davis Scorecard scoring criteria Criterion Score (points) Appearance 2 Color 2 Aroma and bouquet 6 Total acid ity 2 Sweetness 1 Body 1 Flavor 2 Bitterness 1 Astringency 1 General quality 2 Figure 2 3 Monoterpene alcohols and ketones in various wines (Eggers 2005)
41 CHAPTER 3 METHODS Wine Selection Blanc Du Bois wines were selected based on availability and willingness of wineries to participate. Seventeen different wines were initially submitted to the study. Twelve of these were from Texas from 11 different wineries, two were from a winery in Louisiana, and three were from two different wineries in Florida. All wines were made from Blanc Du Bois grapes, and blends were not considered for this study. W ines from vintage years 2006, 2007, and 2008 were used. Sweetness levels varied from semi sweet to slightly sweet to dry. The wines were kept at 55F (13C). Wine Quality Evaluation In February 2009 each wine was entered into a special evaluation session at the Florida State Fair 21st Annual Wine and Grape Juice Competition. Wines were evaluated from uncovered wine gl asses at room temperature by 26 experienced judges. All wines, labeled 1 through 17 were presented simultaneously to all the judges in a nonrandomized fashion. Judges used a modified version of the 20point Davis Scorecard to evaluate color/clarity (12 pts), aroma (1 5 pts), flavor (1 4 pts), balance (1 5 pts), and overall quality (14 pts). The mean score from the 26 judges was considered to be the wines quality rating. Judges also indicated which medal ( double gold, g old, silver, or bronze), if any, w ould have been awarded to a particular wine. Descriptive Analysis Panel Panelist Selection The second component of the study was the DA panel. The study was advertised by email and word of mouth. Twenty five candidates, mostly graduate students and
42 departm ent faculty and staff who had wine tasting experience, participated in a simple screening test to gauge sensory acuity. A triangle test with two different Sauvignon Blanc wines was administered using appropriate sample randomization. Room lighting was dimm ed to mask any color differences, and panelists were asked to identify the different wine as well as write down some aroma and taste descriptors they perceived in each sample. The 16 candidates who correctly completed the triangle test and best described t he wine aromas and flavors were added to the DA panel. Two panelists dropped out during training, before the intensity evaluation sessions took place. Panelist Training Two samples were excluded from the DA panel due to obvious wine stabilization or contam ination defects identified by the judges. They indicated that these flaws were too substantial to allow proper sample evaluation. A third wine was rejected from the DA panel after panelists unanimously agreed that excessive sulfite was irritating their senses and preventing accurate sensory evaluation. For the duration of the training, panelists sat around a large table in a quiet, well lit conference room at approximately 22C. Room temperature samples of approximately 2 U.S. fluid ounces were poured into Libbey Embassy 6.5 ounce wine glasses that were coded by a label at the glass base. The mouth of each glass was covered with a watch glass to allow a headspace to equilibrate and assist panelists with aroma evaluation. Unsalted crackers and water were provided at all times for palate cleansing, and panelists were instructed to expectorate samples into waste vessels. Samples were presented together but evaluated one at a time at the panelists own pace. Each panel training session lasted approximately one hour, with a maximum of six wines presented per day. The complete set of 14 wines was always evaluated within one week.
43 The first step of training was to familiarize the panel with Blanc Du Bois wines while generating a bank of aroma and flavor descriptors for all the wines being evaluated. The first round yielded 111 terms. Descriptors for each wine were written on a whiteboard by the panel leader, and discussion among panelists was encouraged. Subsequently the term bank was trimmed by eliminating redundant descriptors, those that were too general, and those that were mentioned only once or twice by a single panelist. This condensed list yielded 32 descriptors. For the second round panelists were provided with this list and asked to give an approximate intensity score of low, medium, or high if they detected that attribute in a sample. Panelists were asked to stay within this term bank unless they detected a very prominent descriptor absent from the bank. This data was used to further trim the list with the goal of isolating only the key descriptors. At this point in the study, the Wine Aroma Wheel (Noble and others 1984) was introduced to assist panelists in verbalizing the attributes they perceived. Additionally, sev eral week long sessions were devoted to training the panelists to be consistent when identifying different attributes. Attribute references were introduced for two reasons. The first was to ensure that each panelist was correctly discriminating between eas ily confusable aromas: for example, apple versus pear or melon versus peach. Secondly, some general descriptors such as "fruity," "floral," and "spicy, needed to be narrowed down to more specific terms Panelists worked with the panel leader to refine th e references to obtain the best representations of the attributes. A wine aroma kit, Le Nez Du Vin (Lenoir 2006) was employed to provide sensory standards. If panelists reached a consensus agreement that a particular kit aroma was
44 a good representation of the actual attribute, it was established as a standard. Certain aroma standards, such as apple and cut grass, were found to be better rendered by using the actual source of the aroma. In the case of apple, panelists were initially presented with a selection of different apple varieties, with the fruit freshly shredded into a plastic cup. Of Granny Smith, Red Delicious, Gala, and Fuji varieties, Fuji was chosen as having an apple character most similar to that found in the wines, and thus was established as the reference. The third session with the Blanc Du Bois samples was designed to finalize the descriptor list. Descriptors that were too faint to be consistently identified by most panelists were eliminated, as were descriptors that represented defects attributable to winemaking faults (oxidative notes, sulfite, etc.) and not to the grape itself. Practice sessions were held to familiarize the panelists with the final attribute list and the 15point scale. This scale was anchored with zero defined as not detectable, followed by slight, moderate, intense, and extreme, as seen in Figure 31 To provide a quantifiable basis for intensity ratings, sweet, sour, and bitter solutions were created to represent intensit y values of 2, 5, and 10 (Meilgaard, Civille and Carr 2007) An astringent standard was also created to help panelists discriminate between bitter and astringent (Civille and Lyon 1996) The composition of these solutions is displayed in Table 3 2 A t that point in the study p anelists performed a trial run by recording on paper the intensity ratings of each attribute on the list for all 14 wines This session s data was manually entered into Microsoft Excel and then into SAS 9.1 (Cary, NC) An analysis of variance was run on the data to examine how well the panelists were trained at that point in the study. Panelists who were consistently rating
45 an attribute(s ) higher or lower across wine samples relative to the mean were given individual feedback. Wine Attribute Intensity Evaluation The attribute intensity rating sessions were conducted in three consecutive weeks at the University of Florida Sensory Laboratory The facility has 10 individual booths, each equipped with a computer for sensory test administration and data entry. Standards, crackers and water for palate cleansing, and the sweet and sour solutions (ratings 2, 5, 10) for intensity calibration were pr ovided, as seen in Table 32 Data was recorded using Compusense Five software (Compusense Five 4.8 Sensory Analysis Software for Windows Compusense, Guelph, Canada). Panelists analyzed five wines per day on 3 consecutive days in a week, for 3 weeks, ther eby completing the intensity ratings in triplicate. The 15point scale was used, and panelists rated each attribute from the 15descriptor list for each wine, one wine sample at a time. Samples were coded via random 3digit labels on the glass base, and each panelist received a randomized order of presentation of the five wines tested that day. Chemical Analysis Chemical analysi s of the residual sweetness, pH, color, and TA of each wine was performed in triplicate, using the actual bottles of wine that were presented on that particular day at the attribute intensity evaluation sessions. The residual sweetness assay was performed using a Clinitest Analysis Set For Urine Sugar Testing with modified sugar calculation instructions for use with wine sugar measur ement (Presque Isle Wine Cellars, North East, PA 16428). This test measured reducing sugar concentration. For wines with a sugar content less than or equal to
46 1.0%, the test is accurate to 0.1%. For wines 1 to 5%, it is accurate to 1%. For wines containing greater than 5% residual sugar, the test claims accuracy is substantially lower and is indicated to be only approximate. pH measurement was performed on each sample using a Fisher Scientific Accumet Basic AB15 pH Meter with a 13620631 probe. Color ana lysis was performed on each sample using a Beckman Coulter DU 730 Life Science UV/Vis Spectrophotometer set to read absorbance at 420 nm in 10 mm quartz cuvettes. Zoecklein and others (1995) explain that humans perceive color in part due to the wavelengt h of light reaching the eye. Because brown shades are detected primarily in the 400 to 440 nm wavelength range, absorbance measurement at 420 nm is commonly used as an assessment for white wine color. T itratable acidity was measured by titrating 5 ml of wi ne to pH 8.2 using 0.1 N sodium hydroxide. The following formula was used to calculate TA in g/L tartaric acid (Zoecklein and others 1995) : [(mL base) (N base mol/L ) (75. 0 g/mol )] / mL sample This formula work s under the consideration that the molecular weight of tartaric acid is 150 g/mol. It is a diprotic acid, so it takes two equivalents of sodium hydroxide to neutralize it during titration. Thus the value is divided by two to yield 75.0 g/mol. Gas Chromatography Aroma Volatile Analysis Each wine was subjected to a GC MS analysis. Duplicate samples were run from each of two separate bottles (4 samples total) for each different wine. Static Headspace Solid Phase Micro Extraction (HS SPME) was used to collect volatiles The fiber w TM/PDMS StableFlexTM for manual holder, model 57328U (Supelco, Bellefonte, PA). Each extraction was
47 performed with 10 ml of wine in a 40 ml glass vial with a silicone/PTFE septa screw cap. An internal standard was added: 50 L of 21.425 g/mL paracymene (Aldrich, St. Louis, MO) in methanol, and the headspace was flushed with nitrogen. P cymene was chosen as the internal standard for its stability, noncoelution with other volatiles, nondetection in several test samples run without it, and high odor detection threshold ( for GC O analysis) (Bitar and others 2007) A 20 minute room temperatur e equilibration period and a 30 minute room temperature extraction period were used, both with gentle magnetic stir bar stirring. The gas chromatograph was a PerkinElmer Clarus 500 coupled with a PerkinElmer Clarus 500 Quadrupole mass spectrometer (Waltham, MA). The column was a Restek Stabilwax Crossbond Carbowax 60 m length, 0.25 mm inner di ameter (ID), 0.5 m film thickness (df) column (Bellefonte, PA). The GC MS method had a delay time of 0.5 minutes and ended at 40 minutes Scan duration was 0.2 s (m/z range 25300) with an inter scan delay of 0.1 s. Ionization mode was electron impact. The initial GC oven temperature was 40C with a 2.0 minute hold, and the injector port was held at 220C. The temperature was ramped up at 7.0C per minute to 240C, where it was hel d for 9.5 minutes. The carrier gas was helium at 2 ml/min. Mass spectra were taken of the m/z range 25300, and the ionization mode was electron impact. TurboMass 5. 4.2 GC MS Software (Perkin Elmer, Waltham, MA) was used for data acquisition. G as chromatography mass spectrometry identifications were made by analyzing mass spectra data using libraries in TurboMass. Identifications were confirmed by cross referencing linear retention index (LRI) data with published LRI data or by running standards to obtain an LRI match. Peaks were integrated using the software and semi -
48 quantified relative to the concentr ation of the internal standard. These semi quantifications are noted as such due to the fact that only one internal standard was used. Thus, the concentrations of compounds most similar in structure to p cymene (terpenes) would be the most accurate. Other classes of volatiles have different affinit ies for the triphase SPME fiber and should be considered relative to only similar compounds. Gas chromatography olfactory was run on two samples in order to provide a general idea of the key aromaactive volatiles in Blanc Du Bois wines. Two assessors evaluated each sample in duplicate. Wines on opposing sides of principal component one (Figure 46 ) were selected in hopes of obtaining the most accurate representation of the odor active volatil es in all 14 samples. Because each of these wines appeared on the score plot in the area of a different group of volatiles from the load plot, they were deemed to be good representations of the two general types of wines defined by the DA panel. Compound s that had a relatively closely matched sniff from each assessor, as well as an LRI match from GC MS and literature, were considered to be odor active. The same extraction procedure as for the GC MS was performed. The GC unit was an Agilent 6890 running ChromePerfect software (Justice Laboratory Software, Denville, NJ). The column was an Agilent DBWax, 30 m length, 0.32 mm ID, 0.5 thickness The injector temperature was 220C, the initial column temperature was 40C, the temperature ramp was 7C per minute, the final temperature was 240C with a 5 minute end hold time, and the FID detector temperature was 250C. The carrier gas was helium at 2 mL/min, and the effluent was split 1/3 to the FID detector and 2/3 to the olfactometer.
49 Statistical Analysis Sen sory data from Compusense Five (Guelph, Ontario) was transferred to Microsoft Excel in preparation for statistical analysis with SAS 9.1. Two way ANOVA with replication was used to analyze the DA data for attribute intensity differences among wines and for panelist effects such that the classification variables were sample, panelist, and replication. Tukeys Honestly Significantly Different (HSD) Test was used for mean separation. A significant ( p < 0.05) sample x panelist interaction was found for all attr ibutes except grapefruit, lemon, sour, bitter, and astringent. This is a common issue with wine DA panels Noble and Shannon 1987) To test whether the treatments were actually a significant source of variation, an ANOVA was run again using the significant judge x wine interaction mean squares as the error terms (Lawless and Heymann 1998, Stone and Sidel 2004) All attributes were still statistically significant and thus were included in further analyses. For the gas GC MS analysis, the concentration values of all compounds analyzed were averaged across replications. Methanol was not reported since it was the solvent for the int ernal standard, nor was ethanol. Ethanols large concentration in the samples resulted in the detector being overloaded, preventing accurate quantification. PCA was performed on the DA intensity data plus the chemical (color, TA, pH, residual sugar) assays using the princomp procedure with the correlation matrix in SAS. Nor mally, use of the correlation matrix is reserved for PCA with variables possessing different units, as was the case with the DA plus chemical data. For the volatile data, which had uniform units, the correlation matrix was used due to the large differences in concentrations among different types of compounds. The larger variances associated
50 with these volati les caused the data to resemble data having variables with different units. Using the covariance matrix caused SAS to weigh the volatiles with higher co ncentrations more heavily, skewing the PCA plot into distinguishing only the volatiles that were present in very high concentrations (several thousand g/L). Cluster analysis was performed on both the DA data and the volatile data using the cluster procedure in SAS. Several algorithms were tested on the data with similar results, and the average linkage method was selected. This analysis assisted with visualizing relationships among the wines quality scores, their flavor and aroma characteristics, and thei r aroma volatile constituents based on the way they grouped together (Figure 48 ) compared to their positioning on the PCA plots. Table 31. Final descriptor list and corresponding training references Descriptor Standard Applelike Fuji apple Overripe tropical fruit Overripe melon Peachlike Kit Grapefruit Kit Lemon Kit Rose Kit Honey Kit Greenwood/stemmy Fresh grape stems soaked in base wine Phenolic/Rubber New Oak Kit, Le Nez Du Vin Sweet Sucrose solution Sour Citric acid solution Bitter Caf feine solution Astringent Alum solution ND Slight Moderate Intense Extreme 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Figure 3 1 Intensity rating scale used by DA panel
51 Table 32 Intensity calibration references Concentration (%) in water: Intensi ty 2 Concentration (%) in water: Intensity 5 Concentration (%) in water: Intensity 10 Sweet (sucrose) 2.0 0 5.0 0 10.0 0 Sour (citric acid) 0.05 0.08 0.15 Bitter (caffeine) 0.05 0.08 0.15 Astringent (alum) 1.0 0
52 CHAPTER 4 RESULTS AND DISCUSSION Quality Judging The quality scores of the 14 wines ranged from 10.8 to 16.0 (Figure 41) with a mean of 13.7. Table 4 1 shows the mean score of each wine across all 26 judges, as well as the standard deviation of each wine across judges, which ranged from 1.7 to 2.8 with a mean of 2.2. As discussed previously, quality judging is often somewhat subjective, and these quality scores were intended to give a baseline for comparing sensory attributes with groups of higher or lower rated wine samples. Wine 1, which had the highest quality rating (16.0) was a Louisiana wine. The next seven highest quality wines ( 2 8 ) were all from Texas. Wine 12 was the other Louisiana wine. The Florida wines were 9th and 13th, respectively. The standard deviation in each wine was most lik ely due to judge to judge variability and not to differences among the samples each judge received. The wine samples were judged during the same session in the same room using the same type of glassware. S amples were not randomized from panelist to panelis t meaning that judges evaluated them in the same order. There are myriad factors besides differences between wines that can impact the score a wine receives, such as judge preference, setting, and palate exhaustion. Judges are normally able to accurately distinguish which wines are lighter and darker than others, as well as whether or not a wine lacks brilliant clarity. Aroma, flavor, balance, and clarity are more subjective and were not explicitly defined in this quality evaluation, but rather left to the experience of the judges.
53 Descriptive Analysis Term Generation Since no descriptive analysis had been performed on Blanc Du Bois until this study, the only aroma and flavor data was anecdotal. Thus the set of 13 attributes established in this study (Table 3 1) was the first to define Blanc Du Bois flavor and aroma characteristics. Sensory Chemical, and Volatile Correlations The scores and mean separations for the attribute intensity ratings are seen in Table 41 There were significant differences between wines for all attributes at p < 0.05. Panelists tended to use the low end of the 15point scale, with the highest average intensity rating for an attribute topping out at 5.13 for sweetness of wine 6 The most intense attribute across all wines was sourne ss, with an average intensity rating of 2.20, while the least intense was rose, at 0.39. Panelists used the sweet and sour standards to define the number ratings on the intensity scale. The 5% sucrose solution represented an intensity value of 5 on the scale. The two sweetest wines, which measured 4 % and 5.3% residual sugar by weight, were rated 5.13 and 5.06 respectively for sweet intensity, as seen in Table 4 1, indicating that the panelists were quite accurate rating that attribute. Attribute volatile correlations are shown in Table 43; volatiles are presented in order of elution. Wine samples are coded from 1 through 14 with 1 being the highest quality wine and 14 being the lowest. All correlations are at a p value of 0.10. Apple character ranged from 0.87 to 2.03 with a mean of 1.44. Wine 10 had the highest mean, and wine 13 had the lowest. There were two main groupings, with wines 11, 8 9 and 13 having lower means than the other wines. Apple character correlated
54 positively with peach (0.455) and negatively with lemon ( 0.523) greenwood/stemmy ( 0.480) and phenolic /rubber ( 0.588) attributes. Apple did not correlate positively or negatively with any volatiles, although ethyl isobutanoate, ethyl butanoate, and ethyl 2methylbutanoate, compounds often associated with apple aroma, were shown by GC O (Figure 49 ) to be odor active. Because these compounds elute on the GC MS spectra near the ethanol peak, which was large to the point of overloading the detector, it is possible that their calculated concent rations were not very accurate, resulting in no apparent correlation. Ethyl 2methylbutanoate was present at low levels (~10 g/ L) in some wines but was either absent or present below detection limits in others. It is also possible that the panels percept ion of apple character was influenced by other volatiles besides these esters, confounding the correlation. Interestingly, sample 10, which had the highest apple intensity mean, did have the highest ethyl 2methylbutanoate concentration. Ethyl butanoate ( 0.485) and ethyl 2methylbutanoate ( 0.467) had weak negative correlations with quality. Overripe tropical fruit character ranged from 0.33 to 1.86 with a mean of 0.84. Wines 7 6 and 5 had means over 1.00, but only wine 7 was different from the other sam ples. There was not much separation among the wines rated lower than 1.00. Overripe tropical fruit correlated positively with peach (0.543) rose (0.695) honey (0.844) sweet (0.815) and residual sugar (0.468) and negatively with grapefruit ( 0.687) lem on ( 0.657) greenwood/stemmy ( 0.601) phenolic ( 0.561) sour ( 0.762) bitter ( 0.707) and astringent ( 0.501)
55 Overripe tropical fruit correlated positively and negatively with a number of volatiles, as seen in Table 4.2. Of note is the positive correlation with ethyl acetate (0.609) a compound whose solvent or overripe fruit aroma, if above its threshold, is considered to be a wine fault. It results from the esterification of ethanol and acetic acid and occurs when sugar is fermented by wine spoilage yeasts and bacteria a phenomenon that can occur during the process of fruit becoming overripe (Klieber and others 2002) This attribute was also the only one to correlate with the volatile 3methylbutyl octanoate ( isoamyl octanoate) (0.574) a compound that can exhibit chocolatelike, liqueur, or fruity aromas (The Good Scents Company 2010) There was a concern that a dumping effect might occur with the overripe t ropical fruit attribute. Several wines in the study had some oxidative/sherry notes, but the panel reached a consensus that these aromas should not be included in the attribute list, as oxidation character is generally considered to be a defect not intrins ic to the grape. Some panelists noted that overripe tropical fruit character shared some similarities with oxidative notes and thus could be confused, but they were trained to avoid using overripe tropical fruit attribute to describe oxidation character. N evertheless there is a p ossibility it occurred to some extent, as the positive correlation between the overripe tropical fr uit attribute and ethyl lactate (0.626) and diethyl succinate (0.658) concentration implies. As mentioned previously, these compounds in wine can be indicative of oxidative aging. Peachlike character ranged from 0.51 to 2.42 with a mean of 1.56. Wine 6 had the highest peach intensity mean, and wines 9, 13, and 12 had the lowest means. There were three main groups: those below 1.00, four from 1.26 to 1.67, and the rest 1.82 and
56 higher. Peach character correlated positively with apple (0.455) overripe tropical fruit (0.543) rose (0.695) and sweet (0.521) and negatively with lemon ( 0.557) greenwood/stemmy ( 0.776) phenolic ( 0.700) sour ( 0.471) and bitter ( 0.693) Peach also had weak positive correlations with pH (0.476), residual sugar (0.468), and quality (0.462). Peach correlated positively with isoamyl acetate (0.547) which exhibits a banana aroma, and isoamyl hexanoate (0.548) which is described as applelike and fruity (The Good Scents Company 2010) The lactones with which peach character is often associated were not detected in any wine samples. Isoamyl acetate had a posi tive correlation with quality (0.500) Given that peach correlated positively with quality, it appears that isoamyl acetate may be an important component of the peach aroma found in this wine, and thus may be important to the quality of Blanc Du Bois wine. GC O d ata also supported this finding, as it was odor active for both wines analyzed (Figure 4 9 ) Grapefruit character ranged from 0.38 to 1.72 with a mean of 1.14. Wine 2 had the highest grapefruit intensity mean, and wine 7 had the lowest mean. Wine 2 was significantly higher than wines 14 6 and 7 which were all rated below 1.00, and no other wines besides 2 and 7 were significantly different. Grapefruit character correlated positively with lemon (0.626) sour (0.548) bitter (0.665) and astringent (0.532) and negatively with overripe tropical fruit ( 0.687) rose ( 0.457) honey ( 0.643) sweet ( 0.635) sour ( 0.548) and residual sugar ( 0.725) Grapefruit had a weak positive correlation (0.479) with ethyl decanoate and correlated negatively with a number of compounds, especially alcohols. Ethyl decanoate
57 was detected as odor active for both wines when evaluated by GC O but exhibited more of a fruity character according to the sniffers. It is likely that this compound is an element of the grapefrui t character perceived by the panelists and not solely responsible for the aroma. Lemon character ranged from 0.64 to 1.72 with a mean of 1.12. Wine 11 was different from wines 6 and 7 but all other wines were the same. Lemon character correlated positivel y with grapefruit (0.626) sour (0.852) bitter (0.559) and astringent (0.580) and negatively with apple ( 0.523) overripe tropical fruit ( 0.657) peach ( 0.557) rose ( 0.755) honey ( 0.821) sweet ( 0.821) and residual sugar ( 0.784) Lemon correlat ed with volatiles in a similar pattern to grapefruit, with the exception that lemon had a positive correlation with hexanol (0.662) This alcohol is not a prominent component of lemon oil or juice, but it is known to exhibit a green, resiny, or woody aroma (Acree and Arn 2004, The Good Scents Company 2010) Hexanol was found to be odor active during GC O evaluat ion of one of the Florida wines Grapefruit and lemon character were positi vely correlated (0.626), and it is possible that each attribute had a similar set of compounds causing that aromas presence in the wine. This would explain why both grapefruit and lemon paired together as being either high or low from sample to sampl e. The alternative explanation is that the panel was not accurately distinguishing one attribute from the other. Rose character ranged from 0.13 to 0.77, with a mean of 0.39, the lowest of all attributes. Panelists often described this attribute as being very f aint but easily identifiable. Wine 6 had the highest rose character and was different from wines 9, 10, 12, and 13, which all had means of 0.13. All other wines were statistically equivalent,
58 ranging from 0.23 to 0.73. Rose correlated positively with overr ipe tropical fruit (0.695) peach (0.759) honey (0.607) sweet (0.771) pH (0.456) residual sugar (0.717) and quality (0.462) and negatively with grapefruit ( 0.457) lemon ( 0.750) greenwood/stemmy ( 0.654) phenolic ( 0.527) sour ( 0.848) bitter ( 0.781) and astringent ( 0.705) Rose had a positive correlation with isoamyl acetate (0.477) methyl octanoate (0.465) isoamyl hexanoate (0.512) methyl decanoate (0.646) and hexanoic acid (0.551) Of these none are noted in literature of exhibiting specifically rose aroma. Two compounds detected in the wines that are commonly associated with rose aroma are phenethyl alcohol and phenethyl acetate, but neither showed a correlation with rose. The GC O analysis confirmed phenethyl alcohol to be odor active in one of the two wines with the odor being described by the sniffers as floral The lack of a correlation does not necessarily mean they are not at least partly responsible for the rose aroma perceived by the panelists. Since rose was the faintest aroma perceived in most of the wines, the intensity values are low (below 1) and thus wines with the most intense rose aroma are scored similarly in regard to intensity relative to those with little rose aroma. This narrow range may be to blame for the lack of a correlation between these volatiles and rose aroma. Honey character ranged from 0.21 to 2.54 with a mean of 0.97. Wines 7 and 6 separated from all other samples as having higher honey intensity: 2.54 and 2.33, respectively. These were also the two sweetest wines from both a sweetness character and residual sugar level standpoint. The next two highest means, wines 12 and 1 (1.33 and 1.23) were different from 9 10, 13 and 8 which were the four lowest in honey
59 intensity. Honey correlated positively with ov erripe tropical fruit (0.844) rose (0.607) sweet (0.913) and residual sugar (0.875) and negatively with grapefruit ( 0.643) lemon ( 0.821) sour ( 0.839) bitter ( 0.566) and astringent ( 0.594) Honey correlated positively with phenethyl alcohol (0.6 84) which is known to exhibit rose and honey aromas (Acree and Arn 2004) It correlated positively with 17 other volatiles as well, including isoamyl alcohol (brandy, wine, pleasant) (0.761) (The Good Scents Company 2010) ethyl lactate (0.807) acetic acid (0.592) and a number of esters. Honey c orrelated strongly with sweet (0.913) and t he two wines with the most intense honey character were 6 and 7 the sweetest wines. Thus it seems likely that many of the honey volatile correlations are linked to the sweet volatile correlations. It therefore may not be valid to associate all of these correlated volatiles with being responsible for honey character, since they may actually have just been the compounds that happened to occur at high levels in sweet wines and thus associated with the honey attribute. Greenwood/stemmy character ranged from 0.33 to 1.23, with a mean of 0.74. Wines 12, 9, and 13 had the highest intensiti es (all 1.23) and separated themselves from wines 3 7 4 and 6 which had the l owest greenwood/stemmy intensities Greenwood/stemmy character correlated positively with phenolic (0.835) sour (0.465) bitter (0.706) and color (0.593) and negatively with apple ( 0.480) overripe tropical fruit ( 0.609) peach ( 0.776) rose ( 0.653) sweet ( 0.591) residual sugar ( 0.603) and quality ( 0.678) pinene (0.589) and furfural (0.740) pinene is known to contribute a piney, woody, or resinous character,
60 which supports the greenwood/stemmy description panelists used. Furfural, as mentioned previously can form as a breakdown product of plant starches such as xylose or hemicellulose or form from carbonyl amine browning, also known as Maillard reactions By consensus the panelists agreed that none of the wines tasted particularly like they were barrel aged, but since we lacked knowledge of how these wines were made, that source as a furfural contributor cannot be ruled out. There is an outside possibility that wines higher in furfural were not destemmed as carefully as other grapes during the harvesting process. It should be noted that furfural had a negative correlation with quality ( 0.495) ; most of the higher quality wines had lower or unmeasurable concentrations of furfural. Phenolic/rubber character ranged from 0.41 to 3.26, with a mean of 1.27. Wines 9 and 13 had the highest phenolic/rubber intensity and were different than all other wines, with means of 3.26 and 3.21, respectively. Wines 12 and 14 had the next highest means at 1.67 and 1.56 and were different from 3 6 and 4 which had the lowest intensities, at 0.46, 0.44, and 0.41. Phenolic/rubber character correlated positively with g reenwood/stemmy (0.835) and bitter (0.504) and negatively with apple ( 0.588), overripe tropical fruit ( 0.561), peach ( 0.700) rose ( 0.527) and quality ( 0.555) Phenolic/rubber correlated positively with furfural (0.762) but not with any other compounds. It is possi ble that the phenolic character panelists perceived in some wines was due to low threshold phenolic compound(s), such as those discussed previously, which were present below detectable concentrations Sweet character ranged from 0.56 to 5.13, with a mean of 1.87. Wines 6 and 7 had far and away the most sweet intensity, with means of 5.13 and 5.06, respectively,
61 and were different from all other wine samples. Wine 1 also separated from all other samples at 2.97. Wine 4 at 2.23 was different from all samples besides 5 at 1.85. The rest of the samples were rated as having lower sweet intensity, from 1.46 down to 0.56. Sweet correlated positively with overripe tropical fruit (0.815) peach (0.521) rose (0.771) honey (0.913) and sugar (0.962) and negatively wi th grapefruit ( 0.635) lemon ( 0.821) greenwood/stemmy ( 0.591) sour ( 0.869) bitter ( 0.799) and astringent ( 0.781) Sweet correlated positively wi th a large number of volatiles similar in identities to those that correlated with honey character. Ad ditionally, sweet correlated with sabinene hydrate (0.550) and linalool (0.455) two monoterpenes. Sabinene hydrate is known to exhibit cool, minty, and woody aromas, while linalool has a distinct floral aroma. These compounds were probably naturally present in the Blanc Du Bois grape musts, since free terpenes are generally stable throughout fermentation or hydrolyzed from a terpene glycoside. Linalool was found to be odor active for one of the t wo wines when evaluated by GC O. This wine also had a high (5 .13) sweetness intensity rating. Four wines ( 6 7 12, 14) had very high levels of ethyl lactate (>100 g/ L, compared to <20 g/L or none for other samples). These same four also placed in the top five with respect to diethyl succinate concentration. Wines 6 and 7 were the sweet wines rated >5 sweet intensity. Sour character ranged from 1.23 to 2.95, with a mean of 2.20. No single wine separated from the rest, as all samples were less than 0.37 intensity units apart. The sample with the lowest mean, w ine 7, was the same as 4 other wines, and the samples with the highest means ( 3 and 13) were also the same as 4 other wines. It seems
62 plausible that the very low sour rating of wine 7 could be due to its high sweetness (5.06) exerting a masking effect on the panels sourness perception, but this does not appear to be the case. Wine 7s TA is the second lowest at 0.49 mg/L tartaric acid, and the pH is tied for third highest at 3.62, making it one of the least acidic samples and thus supporting the sensory data. Sour correlated positively with grapefruit (0.548) lemon ( 0.852) greenwood/stemmy (0.465) bitter (0.667) astringent (0.749) and TA (0.501) and negatively with overripe tropical fruit ( 0.762) peach ( 0.471) rose ( 0.848) honey ( 0.839) sweet ( 0.869) and residual sugar ( 0.835) Sour correlated positively with only hexanol (0.536) Interestingly and despite the fact that hexanol is not a key aroma volatile in lemons, sensory data correlated it with sourness. It is logical that sourness (a taste sens ation) was not correlated to the volatile acid compounds (e.g. hexanoic acid, octanoic acid), since wine acidity is determined primarily by the content of malic and tartaric acids, which are not volatile. Bitter character ranged from 0.54 to 2.38, with a m ean of 1.37. Wine 11 (2.38) was the same as 13 (1.74), 9 (1.67), and 2 (1.64) but it was different from the rest. Most wines ranged from 1.74 down to 1.04, with the two sweetest wines, 6 and 7 having the lowest bitter intensity ratings of 0.64 and 0.54, suggesting that increased sweetness masked the panelists bitterness perception. Bitter correlated positively with grapefruit (0.665) lemon (0.559) greenwood/stemmy (0.706) phenolic /rubber (0.504) sour (0.667) astringent (0.749) and color (0.580) and negatively with overripe tropical fruit ( 0.707) peach ( 0.693), rose ( 0.781) honey ( 0.566) sweet ( 0.799) residual sugar ( 0.809) and quality ( 0.505)
63 Bitter correlated positively with furfural (0.543) and butyrolactone (0.565) Furfural is known to smell of caramel or bitter almond (The Good Scents Company 2010) but it is unclear whether either of these compounds contributes to bitterness on the tongue. Astringent character ranged from 0.21 to 1.33, with a mean of 0.74. Similar to the sour attribute, there were no standout astringent wines or groupings. As with the least bitter wines, the two least astringent wines were also the sweetest, indicating a possible perception masking effect. Astringent correlated positively with grapefruit (0.532) lemon (0.580) sour (0.750) and bitter (0.776) and negatively with overripe tropical fruit ( 0.501) rose ( 0.705) honey ( 0.594) sweet ( 0.781) and residual sugar ( 0.741) Astringent correlated positiv ely with only butyrolactone (0.652) al though no studies have found this compound to exhibit an astringent effect in wines. Normally it is perceived as a buttery aroma (Maarse 1991) No wines contained this volatile at concentrations approaching the 35 g/L threshold value reported by literature (Selli and others 2008) Astringency is a mouthfeel attribute and generally produced by nonvolatile components in the liquid phase. Phenolics and tannins are usually associated with this sensory attribute. The volatile 3 methyl 1 pentanol correlated with more sensory attributes than any other volatile in the study. It correlated positively with overripe tropical fruit (0.802) rose (0.671) honey (0.891) sweet (0.928) and sugar (0.900) and negatively with grapefruit ( 0.652) lemon ( 0.720) greenwood/stemmy ( 0.447) sour ( 0.772) bitter ( 0.706) and astringent ( 0.747) It is evident that this volatile was higher in the sweeter wines and thus the attributes those wines tended to exhibit. Residual sugar content, however, does not seem to be requisite to its formation, as wines 2 and 14, which were finished
64 relatively dry both contained 3methyl 1 pentanol Besides these and the sweet w ines 6 and 7 only wine 1 contained this volatile. Some wine sample chromatograms contained 2 peaks that eluted closely together around retention times 19.04 and 19.12, or LRI 1566 and 1571. Both mass spectra analyses returned good matches for linalool, wi th the first one being a slightly higher match A linalool standard was run and had a calculated LRI of 1567. As seen in Table A 1, some samples only had a peak for the first LRI (1566), termed linalool, others only had a peak for the second LRI (1571), termed linalool2, and some had both. The identity of these compounds was not confirmed, but linalool oxides were ruled out due to their approximately 100point lower literature LRI values. Chemical Analysi s Color (spectrophotomet ric absorbance at 420 nm) ranged from 0.048 to 0.203 ( Figure 42 ) with a mean of 0.100 across all wines. Wine 11 ( a bsorbance 0.203) had significantly higher absorbance than all others, and wine 14 had the next highest mean ( 0 .140) and was different from all other wines except 10 ( 0 .1 34). Wine 4 was significantly lighter than all others. The rest of the wines were closer to the mean, ranging from 0.068 ( wine 2 ) to 0.140 ( wine 14). Color correlated positively with greenwood/stemmy (0.593) and bitter ( 0.580) and had a negative correlation with quality ( 0.621) Color correlated positively with a number of volatiles, including several terpenes, several esters, and furfural. The terpenes included pinene (0.634) terpinolene (0.501) and nerol oxide (0.676) These volatiles are not noted in literature to contribute to browning. White wine color is influenced by a number of factors, including the amount
65 of time juice spends in the pressed but unsulfited state, the amount of phenolic compounds in the juice, as well as the juices contact ti me with the skins, which can influence the concentration of phenolic compounds in the wine. Phenolic compounds such as catechins and epicatechins are susceptible to oxidation, which results in visual browning or yellowing of the wine over time ( Labrouche and others 2005). T itratable acidity ranged from 0.44 to 0.70 g tartaric acid/L (Figure 43) with a mean of 0.58 g tartaric acid/L. Partly due to this small range, the Tukey groupings for the mean separation did not reveal any single wine or small group as being higher or lower than others in acidity. Wine 14 had the lowest mean and was statistically the same as 3 other wines. TA had a positive correlation with sour (0.501) It had a negative correlation with pH ( 0.543) which is logical considering that a wine with more acid (more TA ) has more free hydrogen ions, prompting a lower pH value. pH ranged from 3.28 to 3.94 (Figure 44) with a mean of 3.50. More than half of the wines belonged to one Tuk ey grouping ( d; 9 5 13, 4 3 10, 13, 12, 1 ), ranging from 3.47 to 3.28, respectively as seen in Table 42 Wine 14 had the highest mean and was different from all other wines except 8 (3.75). pH correlated positively with peach (0.476) and rose (0.456) and negatively with TA ( 0.543) Because wine acidity is pr edominantly a function of malic and tartaric acid concentration, it is improbable that the volatiles that correlated with peach and rose character had much influence on pH. It is possible, however, that pH could affect fermentation factors and yeast metabolism, leading to differing levels of those volatiles. It has been shown that ester hydrolysis occurs during wine aging and that the reaction rate is dependent on both temperature and hydrogen ion concentration (Ram ey
66 and Ough 1980) The study by Ramey and Ough (1980) also showed that acetate esters hydrolyze more quickly than ethyl esters. Thus it raises the question of whether the Blanc Du Bois wines with high peach character had more isoamyl acetate produced init ially during fermentation or lost less isoamyl acetate to acid catalyzed hydrolysis during storage. In the latter case, the highest pH wines would be expected to have higher ester concentrations due to slower acid catalyzed hydrolysis and that may have been the case for some of these samples given the positive pH peach/rose correlation. Additionally, in this case a link between age and ester content may exist, but validation would require a larger sample set from each vintage in this study (2006, 2007, 2008). Another possibility is that the higher pH wines coincidentally had higher levels of sulfite (SO2). It has been shown that more heavily sulfated wines retain their volatile ester and volatile alcohol levels better than nonsulfated wines (GardeCerdn and Ancn Azpilicueta 2007) This explanation seems less plausible since higher SO2 levels can lighten wine color, and there was no color pH correlation for the wines in this study. Residual sugar ranged from 0.0 to 5.3% (Figure 45) with a mean of 1.4%. Wine 7 had the highest residual sugar mean and separated by itself, as did wine 6 at 4.0%. Wines 4 (2.0%), 1 (1.7%), and 5 (1.6%) had the next 3 highest means, al though the latter two were statistically the same as those wines with means 0.9 and higher. Wines with residual sugar ranging from 0.0 to 0.9% were all statistically equivalent. Residual sugar correlated positively with overripe tropical fruit (0.849) peach (0.468) rose (0.717) honey (0.875) and sw eet (0.962) and negatively with grapefruit ( 0.725) lemon ( 0.784) greenwood/stemmy ( 0.603) sour ( 0.835) bitter ( 0.809) and astringent ( 0.741)
67 Residual sugar correlated positively with a number of esters and alcohols and negatively with nothing. It is unlikely that there is much to be inferred from volatilesugar correlations, as the sweeter wines were probably fermented, stabilized, and back sweetened to achieve the desired residual sugar content. The other possibility is that some nonfermentabl e reducing sugars were present and imparted a slight sweetness to some wines. Neither scenario should have had any influence on volatile profiles, nor would volatiles have influenced the amount of residual sugar in a wine. Principal Component and Cluster A nalyse s Principal Component Analysis: DA and Chemical Data PCA was conducted on all attributes and chemical measurements since the ANOVA determined them all to be significantly different. As seen in Figure 46, principal component 1 (PC1) explained 53.74% of the variation observed in the attribute intensity data, and principal component 2 (PC2) explained 13.53% of the variation, for a total of 67.27% of the datas variation explained by the biplot. In general the PCA load plot confirmed what was observed in the correlation analysis. Attributes found on one side of PC1 of the load plot (Figure 46) correlated positively with those around them and negatively with those on the opposite side of the axis. Sour, lemon, astringent, bitter, and grapefruit were oppos ite (approximately 180 degrees apart) from residual sugar, sweet, overripe tropical fruit, apple, rose, and peach. PCA can display both the attr ibutes that are closely related as well as where the samples fall relative to the variables most responsible for their differences. In this study, the PCA showed which wine samples grouped with which attributes. The grouping of w ines 1, 5, and 6 on the score plot (Figure 47) appeared to be driven by apple character, and their apple intensity ratings were 46th, res pectively, as
68 seen in Table 41. Additionally, wine 6 was placed farther out near the sweet and residual sugar variables. The DA intensity data for residual sugar and sweet supported this, as wine 6 was one of the two sweetest wines. According to the PCA, wine 4 was driven by peach character and the DA data showed peach intensity to be fourth highest among all wines at 1.94. Wine 3 was driven by high TA and measured 0.70 g tartaric acid/L, the highest among all wines. Wines 2, 8, and 11 also appeared to be heavily influenced by TA, as that was the only variable in that area common to these samples. Supporting this observation was the fact that the intensity data rated each sample highly with respect to the other wines. Wine 10 appeared to be influenced by a number of attributes in that quadrant of the biplot. Analysis of the intensity ratings indicated that it had the highest lemon rating of all wines and was rated somewhere in the middle for the other attributes in that area of the biplot, such as TA, grapefruit, sour, and astringent. Wines 9, 12, and 13 were plotted in the quadrant with bitter, greenwood/stemmy, phenolic, and color attributes (Figures 4 6, 4 7 ). Wine 12 was highly correlated with color and also had the highest mean absorbance measurement : 0 .203. Wine 9, a Florida wine, was closer to the phenolic and greenwood/stemmy attributes, and it had the highest phenolic/rubber intensity rating, at 3.26, although wine 13, the other Florida wine, was very close at 3.20. Wine 13 was placed lower on the bi plot than wine 9, which implied that it could have been more influenced by bitter character than wine 9. Wine 13 ranked secondhighest in bitterness intensity at 1.74, slightly higher than wine 9 at 1.67. Wines 14 and 7 were placed in the upper right quadr ant (Figure 47 ). Wine 14 appeared to be driven by pH but it placed near the color attribute on PC2 as well.
69 Supporting this data is the fact that it had the highest pH (3.94) and seconddarkest color (0.140). Wine 7 was placed directly in line with honey character, and it did have the highes t honey intensity of all wines: 2.54. Given the very strong correlation between honey and sweet noted previously, it is logical to infer that wine 7 was also driven by sweetness. Cluster Analysis: DA and Chemical Data M eilgaard and others stated that  cluster analysis identifies groups of observations based on the degree of similarity among their ratings (Meilgaard, Civille and Carr 2007) Cluster analysis identif ied groups of wine samples according to how similar their sensory intensity ratings were. Although samples may be a similar distance apart from one another on a PCA biplot, it does not imply that they have the same degree of difference in terms of what att ributes are driving their placement. For example, in the DA PCA (Figures 4 6, 4 7 ) wine 12 was closer to wine 9 than to wine 14, but cluster analysis showed wine 12 to be more closely related to wine 14. As seen in the cluster analysis ( Figure 48 ) wines 6 and 7, the sweetest wines, were least related to the other samples and branched off near the top of the chart. The next wines to group together were wines 9 and 13, the Florida wines, which were high in citrus and greenwood/stemmy character. Beyond thes e clusters the wines were more closely related, as shown by the shorter vertical distances between branches. Wines 12 and 14 clustered together, likely a result of their dark color. A group of high quality wines, 1, 4, and 5, clustered together; this was probably a result of their higher intensity peach character. Finally, a large and varied group of wines clustered together 2, 3, 8, 10, and 11. Judging from the PCA plot, these wines were driven by a number of variables, especially TA
70 Principal Component Analysis: Volatile Data Due to t he large number of volatiles (60+ ) identified in the wines, it was difficult to pinpoint whether or not a particular volatile was influencing a samples positioning on the PCA plot. It was more useful to examine the placement of the samples on the plot relative to groups of volatiles and look to see if those volatiles were related to each other from the standpoi nt of their molecular composition. S imilar ly structured volatiles often exhibit similar aromas, as is the case with esters often being perceived as fruity and terpenes being perceived as woody or spicy. Some trace volatiles that were detected in only one or two wines were excluded from the PCA and cluster analyses. The rest are shown in Table 51. For this PCA, PC1 ( 26.74%) and PC2 (22.40%) combined to explain 49.14% of the variability in the data, as seen in Figures 49 and 410. Many volatiles did not place very close to either principal component. This was somewhat expected given the large number of variables; it i s more difficult to explain the variability of so much data with just two principal components. Additionally, many volatiles did not differentiate the wines, so their proximity to a sample might be coincidental. Wine 6 was the most unique of all the samples; it was isolated in its quadrant far from the other samples. The volatiles in this area were terpenes such as linalool (31) and sabinene hydrate (25), and esters such as hexyl acetate (13), isoamyl hexanoate (24), 3 methylbutyl octanoate (37), and methyl octanoate (20). This wines location on the biplot was likely driven by linalool, as it had by far the highest concentration among all samples at 44.0 g/L the next closest was 16.3 g/L. A group of high quality wines 2, 3, and 5 were relatively clo se to each other in the upper left quadrant. Wine 5 was close to butanol (10) and pcymene (14) and not
71 much else. P cymene was not considered to be a source of variability since as the internal standard it was input as an equal concentration in all samples (107.1 g/L). Butanol was not odor active. Wine 10 was also in this area and fell directly on PC1, implying a possible relationship with phenethyl acetate (41), but the GC MS analysis did not identify this volatile in the sample. It was, however, relativ ely close to hexanol (19), a volatile whose highest concentration was found in wine 10 (69.7 g/L, next closest was 45.8 g/L). Wine 2 was closer to PC2, where nearby volatiles consisted of a number of esters and organic acids, including ethyl decanoate (36), ethyl octanoate (22), ethyl dodecanoate (42), methyl decanoate (35), octanoic acid (45), and hexanoic acid (43). Wine 2 placed in the top three of all wines for each of those six compounds concentrations and probably would have been located directly i n the midst of each of them if it were not for its very high level of phenethyl acetate (20.1 g/L), a volatile located on PC1. Wines 9 and 13, the Florida wines, were placed very close together along PC2, as were wines 11 and 14. The only nearby volatiles belonging to PC2 were butyrolactone (38), furfural (27), E 2 hexenol (21) and linalool2 (32). Wines 1, 12, and 7 were also in the bottom right quadrant, but their variability was better explained by PC1. There were many volatiles clustered in this area, i ncluding esters, alcohols, terpenes, aldehydes, and organic acids. The odor active compounds in area were phenethyl alcohol (44), ethyl hexanoate (12), and acetic acid (23), as well as the ethyl esters ethyl butanoate (3) and ethyl 2methylbutanoate (4) sl ightly above PC1 in the upper right quadrant.
72 It was evident from studying the PCA of the volatiles that less inference could be drawn from this data as compared to the PCA of the DA data. Most of the volatiles found in the samples, such as octanoic acid ( 45) and hexanoic acid (43), did not directly cont ribute to an aroma in the wine. These volatiles have aroma thresholds substantially higher than the levels found in these wine samples, yet differences in their concentrations might still influence the confi guration of the samples on the biplot. In some cases nonodor active volatiles that were present in relatively high concentrations in a wine may have caused that sample on the score plot to place far away from a low concentration aroma active volatile that did influence the aromatic character of that wine. Another explanation is that due to the large number of variables on the load plot, there were many possible rotations for explaining approximately 49% of the variability in the dataset For example, wine 5 placed in the middle of the upper left quadrant, between PC1 and PC2. It appears to be correlated with butanol (10), which is its most nearby attribute. Instead, its placement was likely dictated by its high levels of methyl decanoate (11.0 g/L, next cl osest was 7.2 g/L) and phenethyl acetate (16.5 g/L, second highest among all wines). Methyl decanoate (35) was on PC2, and phenethyl acetate (41) was on PC1; therefore, wine 5 was placed in between those two and coincidentally close to butanol, despite t he fact that it did not have a high concentration of th at compound (tied for 7th highest). Perhaps a different rotation that explained slightly less variability in the data may have looked quite different and placed these compounds closer to one another and others farther apart more accurately
73 representing the volatile drivers for wine 5 but at the same time causing other wines interpretations to become less clear. Cluster Analysis: Volatile Data Wines with similar volatile concentration profiles tended t o group together on the volatile cluster analysis. As shown on the volatile data PCA, it was difficult to predict wine quality by examining specific volatiles. Figure 411 does not appear to reveal any quality trends, since clusters grouped both high and l ow quality wines together. This does not guarantee that the volatile data does not influence the quality of the wine but instead indicates that studying the full GC MS derived volatile profiles is probably a poor way to predict Blanc Du Bois wine quality. In some cases, such as the cluster of wines 1 and 11, it appears that the grouping may have been driven by a set of shared volatiles between those wines. As seen in the data in Table A 1, some volatiles were present (or at least measurable) in only two or three wines. It is possible that two wines sharing several of these rare volatiles would cause them to cluster together. The two Florida wines shared similar volatile concentration profiles, and this was highlighted in the cluster analysis, where they gr ouped closely together. Volatile Content: Similarities to Other Wine Styles As discussed previously, esters are a major of component of the general bouquet associated with white wines. This study found that Blanc Du Bois in this regard shares many simila rities with other wine styles noted anecdotally to have similar aroma/flavor profiles. Most of the volatile esters identified as odor active in this study were found to be present in Sauvignon Blanc and odor active in Gewrztraminer and Riesling wines in separate studies (Komes, Ulrich and Lovric 2006, King and others 2008, Guth 1997)
74 The exceptions were methyl octanoate, ethyl decanoate, and octanoic acid, which were not found in the Gewrztraminer study, but were found in this and the Riesling study. Additionally, there were at least 20 odor active volatiles (including cis rose oxide, 2 methoxyphenol, trans ethyl cinnamate, eugenol, 3ethylphenol, vinylguiacol, damascenone) in the Gewrztraminer wines that were not present in Blanc Du Bois. There were at least 15 in the Riesling (including benzaldehyde, isobutyric acid, benzeneacetaldehyde, N (3 methylbutyl) acetamide, damascenone, diethyl malate, 4vinylguai acol methyl vanillate) wines that were not present in Blanc Du Bois. This could have been due to the extraction procedure (SPME) used in this study. The two GC O studies utilized a liquid liquid extraction (Komes, Ulrich and Lovric 2006, Guth 1997) t hat may have extracted higher levels of volatiles, which could have result ed in more of them being present in concentrations sufficient to elicit a response from the human sniffers. If this were not the cas e, it may simply have meant that the wines in this study, and in particular those two chosen for the Blanc Du Bois GC O analysis, had less diverse volatile profiles compared to those of other wine varieties.
75 Table 41 DA attribute intensity, chemical, and quality means with Tukeys HSD mean separation1. Wine letter represents quality rank, with A = highest and N = lowest Attribute Wines: 1 2 3 4 5 6 7 Apple like 1.67 abc 1.36 abc 1.49 abc 1.86 ab 1.62 abc 1.51 abc 1.46 abc Overripe Trop ical Fruit 0.64 bc 0.74 bc 0.85 bc 0.67 cd 1.08 abc 1.46 ab 1.86 a Peach like 1.67 abc 1.28 bcd 2.15 ab 1.94 ab 1.95 ab 2.42 a 1.90 ab Grapefruit 1.28 ab 1.72 a 1.26 ab 1.15 abc 1.08 abc 0.82 bc 0.38 c Lemon 0.87 ab 1.31 ab 1.23 ab 0.97 ab 0.92 ab 0.64 b 0.64 b Rose 0.49 abc 0.44 abc 0.23 bc 0.44 abc 0.73 ab 0.77 a 0.69 ab Honey 1.23 b 0.64 bcd 0.59 bcd 0.95 bcd 0.67 bcd 2.33 a 2.54 a Greenwood / Stemmy 0.77 ab 0.56 ab 0.49 b 0.36 b 0.62 ab 0.33 b 0.41 b Phenolic / Rubber 1.13 bc 0.79 bc 0.46 c 0.41 c 1.03 bc 0.44 c 0.87 bc Sweet 2.97 b 1.46 de 1.08 efg 2.23 c 1.85 cd 5.13 a 5.06 a Sour 1.90 cdef 2.19 bcd 2.95 a 1.95 bcdef 1.72 def 1.36 ef 1.23 f Bitter 1.10 bcd 1.64 ab 1.54 b 1.10 bcd 1.04 bcd 0.64 cd 0.54 d Astringent 0.46 cde 0.64 bcde 1.33 a 0 .67 bcde 0.69 bcde 0.21 e 0.31 de Color (Abs. @ 420nm) 0.103 de 0.068 gf 0.074 f 0.048 g 0.094 def 0.090 def 0.071 gf TA (g/L tartaric acid) 6.2 abcd 6.1 abcd 7.0 a 5.5 bcde 5.4 cdef 6.3 abcd 4.9 ef pH 3.28 e 3.58 bcd 3.37 de 3.42 cde 3.47 cde 3.62 bc 3 .62 bc Residual Sugar ( % weight) 1.7 cd 0.9 def 0.7 ef 2.0 c 1.6 cd 4.0 b 5.3 a Quality Rating 16.0 a 15.5 ab 15.3 ab 14.7 abc 14.5 abc 14.4 abc 13.8 b cd Standard Dev. (Quality) 2.1 1.8 2.6 2.5 2.3 2.2 2.7 1Wines sharing like letters in the mean separation for a certain attribute or chemical measurement are not different in that attributes intensity or chemical property at p < 0.10.
76 Table 4 1 Continued Attribute Wines: 8 9 10 11 12 13 14 Apple like 1.05 bc 0.97 bc 2.03 a 1.06 bc 1.49 abc 0.87 c 1.76 abc Overripe Tropical Fruit 0.51 c 0.36 c 0.97 bc 0.75 bc 0.72 cd 0.33 c 0.82 bc Peach like 1.82 ab 0.77 cd 1.26 bcd 1.33 bcd 0.51 d 0.62 d 2.27 ab Grapefruit 1.38 ab 1.02 abc 1.00 abc 1.28 ab 1.51 a b 1.21 abc 0.85 bc Lemon 1.46 ab 1.26 ab 1.32 ab 1.72 a 1.15 ab 1.38 ab 0.85 ab Rose 0.36 abc 0.12 c 0.12 c 0.25 bc 0.12 c 0.12 c 0.51 abc Honey 0.21 d 0.41 cd 0.92 bcd 0.36 cd 1.33 b 0.33 cd 1.05 bc Greenwood / Stemmy 0.56 ab 1 .23 a 0.64 ab 1.00 ab 1.23 a 1.23 a 0.95 ab Phenolic / Rubber 1.15 bc 3.26 a 0.97 bc 0.89 bc 1.67 b 3.20 a 1.56 b Sweet 0.69 fg 1.02 efg 1.08 efg 0.75 fg 1.03 efg 0.56 g 1.23 def Sour 2.69 ab 2.51 abc 2.53 abc 2.56 abc 2.18 bcd 2.95 a 2.03 bcde Bitter 1.49 b 1.67 ab 1.42 bc 1.25 bcd 2.38 a 1.74 ab 1.56 b Astringent 0.92 abc 0.72 bcde 0.85 abcd 0.78 abcde 1.21 ab 0.77 abcde 0.69 bcde Color (Abs. @ 420nm) 0.072 gf 0.107 dc 0.134 bc 0.093 def 0.203 a 0.081 e f 0.140 b TA (g/L tartaric acid) 5. 9 bcde 6.4 abc 6.5 ab 5.9 bcde 5.3 def 6.0 bcd 4.4 f pH 3.75 ab 3.47 cde 3.29 e 3.37 de 3.31 e 3.45 cde 3.94 a Residual Sugar (% weight) 0.1 f 1.0 de 0.9 def 0.1 f 0.2 ef 0.1 f 0.4 ef Quality Rating 13 .7 bcd 13.3 cd 12.2 de 12.2 de 11.0 e 10.9 e 10.8 e Standard Dev. (Quality) 2.3 2.8 2.1 1.7 2.3 1.8 1.9 1Wines sharing like letters in the mean separation for a certain attribute or chemical measurement are not different in that attribute's intensity or chemical property at p < 0.10.
77 Table 42. DA, chemical and quality correlations significant at p < 0.10 Apple Overripe Tropical Fruit Peach Grapefruit Lemon Rose Honey Green wood / Stemmy Phenolic Applelike 1.000 0.455 0.523 0.480 0.588 Overripe Tropical Fruit 1.000 0.543 0.687 0.657 0.695 0.844 0.609 0.561 Peach like 0.455 0.543 1.000 0.557 0.759 0.776 0.700 Grapefruit 0.687 1.000 0.626 0.457 0.643 Lemon 0.523 0.657 0.557 0.626 1.000 0.750 0. 821 Rose 0.695 0.759 0.457 0.755 1.000 0.607 0.653 0.527 Honey 0.844 0.643 0.821 0.607 1.000 0.423 0.368 Greenwood / Stemmy 0.480 0.601 0.776 0.654 1.000 0.835 Phenolic / Rubber 0.588 0.561 0.700 0.527 0.835 1.000 Sweet 0.815 0.521 0.635 0.821 0.771 0.913 0.591 Sour 0.762 0.471 0.548 0.852 0.848 0.839 0.465 Bitter 0.707 0.693 0.665 0.559 0.781 0.566 0.706 0.504 Astringent 0.501 0.532 0.580 0.705 0.594 Color 0.593 TA pH 0.476 0.456 Residual Sugar 0.849 0.468 0.725 0.784 0.717 0.875 0.603 Quality Rating 0.462 0.462 0.678 0.555
78 Table 42 Continued Sweet Sour Bitter Astringent Color TA pH Sugar Quality App le like Overripe Tropical Fruit 0.815 0.762 0.707 0.501 0.849 Peach like 0.521 0.471 0.693 0.476 0.468 0.462 Grapefruit 0.635 0.548 0.665 0.532 0.725 Lemon 0.821 0.852 0.559 0.580 0.784 Rose 0.7 71 0.848 0.781 0.705 0.456 0.717 0.462 Honey 0.913 0.839 0.566 0.594 0.875 Greenwood / Stemmy 0.591 0.465 0.706 0.593 0.603 0.678 Phenolic / Rubber 0.504 0.555 Sweet 1.000 0.869 0.799 0.781 0.962 S our 0.869 1.000 0.750 0.501 0.835 Bitter 0.799 0.667 1.000 0.776 0.580 0.809 0.505 Astringent 0.781 0.749 0.776 1.000 0.741 Color 0.580 1.000 0.621 TA 0.501 1.000 0.543 pH 0.543 1.000 Residual Sugar 0.962 0.835 0.809 0.741 1.000 Quality Rating 0.505 0.621 1.000
79 Table 43 DA attribute and volatile correlations significant at p < 0.10 ethyl acetate ethyl isobutanoate ethyl butanoate ethyl 2methylbutano ate ethyl 3methylbutanoate hexanal isobutanol pinene isoamyl acetate butanol isoamyl alcohol Applelike Overripe Tropical Fruit 0.609 0.727 0.725 Peach like 0.488 0.514 0.547 Grapefruit 0.667 0.704 Lemon 0.723 0.468 0.509 Rose 0.477 Honey 0.876 0.637 0.506 0.722 0.761 Greenwood / Stemmy 0.589 0.634 0.447 Phenolic / Rubber 0.511 Swe et 0.714 0.508 0.542 0.639 Sour 0.709 0.473 Bitter Astringent 0.482 Color 0.589 0.519 0.653 0.526 TA pH 0.463 Resi dual Sugar 0.681 0.446 0.626 0.720 Quality Rating 0.485 0.467 0.500
80 Table 43 Continued ethyl hexanoate hexyl acetate terpinolene 3 methyl 1 pentanol ethyl heptanoate ethyl lactate hexanol methyl octanoate E 2 hexenol ethyl octanoate acetic acid Applelike Overripe Tropical Fruit 0.802 0.626 0.512 Peach like Grapefruit 0.652 0.508 Lemon 0.720 0.619 0.662 0.513 Rose 0.671 0.466 0.465 Honey 0.891 0.807 0.604 0.497 0.592 Greenwood / Stemmy 0.447 0.488 Phenolic / Rubber 0.618 0.634 Sweet 0.928 0.552 Sour 0.772 0.607 0. 536 0.586 0.474 Bitter 0.706 Astringent 0.747 0.490 Color 0.501 0.470 0.519 0.510 TA 0.592 0.481 0.569 pH 0.503 Residual Sugar 0.900 0.518 Quality Rating 0.482
81 Table 43. Continued isoamyl hexanoate sabinene hydrate nerol oxide furfural decanal ethyl sorbate ethyl nonanoate linalool linalool2 vitispirane octanol Applelike 0.598 Overripe Tropical Fruit 0.587 0.820 0.617 0.493 Peach like 0.548 0.475 0.550 Grapefruit 0.695 0.451 0.652 Lemon 0.655 Rose 0.512 0.515 Honey 0.500 0.556 0.729 0.620 0.500 Greenwood / Stemmy 0.691 0.740 0.677 0.521 Phenolic / Rubber 0.534 0.762 0.522 Sweet 0.602 0.550 0.496 0.751 0.455 0.533 Sour 0.595 Bitter 0.490 0.543 0.735 0.65 7 0.467 Astringent 0.449 0.515 0.499 Color 0.676 0.534 0.853 TA 0.468 pH Residual Sugar 0.550 0.811 0.544 0.644 Quality Rating 0.495 0.514
82 Table 4 3 Continued methyl decanoate ethyl decanoate 3 methylbutyl octanoate butyrolactone ethyl succinate ethyl9 decenoate phenethyl acetate ethyl dodecanoate hexanoic acid phenethyl alcohol octanoic acid Applelike F46 Ov erripe Tropical Fruit F47 0.574 0.658 0.691 0.671 Peach like F48 Grapefruit F49 0.479 0.615 0.649 0.531 Lemon F50 0.580 0.471 0.544 Rose F51 0.646 0.679 0.551 0.490 Honey F52 0.797 0.674 0.684 Greenwood / Stemmy F53 0.529 Phenolic / Rubber F54 0.447 0.535 0.539 Sweet F55 0.684 0.633 0.529 0.649 Sour F56 0.595 0.524 0.538 0.541 Bitter F57 0.565 0.489 0.454 Astringent F58 0.652 0.446 0.590 Color F59 0.528 TA F60 pH F61 0.501 Residual Sugar F62 0.724 0.759 0.729 Quality Rating F63 0.514 0.476 0.449
83 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 A B C D E F G H I J K L M N Wine Sample Quality Score Quality Score Figure 41 Quality scores of wine samples as determined by expert judging panel in decreasing order
84 0.000 0.050 0.100 0.150 0.200 0.250 A B C D E F G H I J K L M N Wine Sample Absorbance @ 420 nm Absorbance Figure 42 Color measured by a spectrophotometer reading absorbance at the 420 nm wavelength. Samples sorted by decreasing quality score
85 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 A B C D E F G H I J K L M N Wine Sample Titratable Acidity (g/L tartaric acid) Titratable Acidity Figure 43 TA measured in grams of tartaric acid per liter. Samples sorted by decreasing quality score
86 3.00 3.10 3.20 3.30 3.40 3.50 3.60 3.70 3.80 3.90 4.00 A B C D E F G H I J K L M N Wine Sample pH pH Figure 44 pH of wine samples. Samples sorted by decreasing quality score
87 0.0 1.0 2.0 3.0 4.0 5.0 6.0 A B C D E F G H I J K L M N Wine Sample Residual Sugar (%) Residual Sugar Figure 45 Residua l sugar as percent weight of wine samples. Samples sorted by decreasing quality score
88 Figure 46 PCA variables plot showing PC1 and PC2 for the DA attribute intensity data
89 Figure 47. PCA samples plot showing PC1 and PC2 for the DA attribute intens ity data. Numbers indicate quality ranking of the wine, with 1 being highest quality
90 F igure 48 Cluster analysis for the DA attribute intensity data. Numbers indicate quality ranking of the wine, with 1 being highest quality
91 Figure 49 PCA variables plot showing PC1 and PC2 for the GC MS volatile data. Volatiles determined to be odor active using GC O are shaded in. See Table 5 1 for cross reference of volatiles
92 Figure 41 0 PCA samples plot showing PC1 and PC2 for the GC MS volatile data. Numb ers indicate quality ranking of the wine, with 1 being highest quality
93 Figure 41 1 Cluster analysis for the GC MS volatile data. Numbers indicate quality ranking of the wine, with 1 being highest quality
94 Table 51. Key for identification of volati les used in PCA on Figure 49 plus L inear R etention I ndex values for volatiles Number Volatile Identity LRI 1 ethyl acetate 902 2 ethyl isobutanoate 984 3 ethyl butanoate 1056 4 ethyl 2 methylbutanoate 1068 5 ethyl 3methylbutanoate 1085 6 hexanal 1 096 7 isobutanol 1110 8 pinene 1124 9 isoamyl acetate 1140 10 butanol 1161 11 isoamyl alcohol 1223 12 ethyl hexanoate 1252 13 hexyl acetate 1291 14 p cymene 1294 15 terpinolene 1304 16 3 methyl 1 pentanol 1347 17 ethyl heptanoate 1353 18 et hyl lactate 1370 19 hexanol 1372 20 methyl octanoate 1412 21 E 2 hexenol 1434 22 ethyl octanoate 1461 23 acetic acid 1477 24 isoamyl hexanoate 1481 25 sabinene hydrate 1499 26 nerol oxide 1502 27 furfural 1508 28 decanal 1531 29 ethyl sorbate 1541 30 ethyl nonanoate 1559 31 linalool 1569 32 linalool2 1571 33 vitispirane 1573 34 octanol 1579 35 methyl decanoate 1622 36 ethyl decanoate 1663 37 3 methylbutyl octanoate 1684 38 butyrolactone 1699 39 ethyl succinate 1706 40 ethyl9 decenoate 1717 41 phenethyl acetate 1861 42 ethyl dodecanoate 1866 43 hexanoic acid 1870 44 phenethyl alcohol 1962 45 octanoic acid 2085
95 CHAPTER 5 CONCLUSION The objective of this study was to characterize Blanc Du Bois wine quality, sensory attributes, flav or volatiles, and the relationships among these. There were differences in wine quality as assessed by the quality judging panel, with 14 wines earning scores ranging from 10.8 to 16.0 on a 20point scale. An attribute list representing 13 prominent aromas or flavors perceived in the wines was agreed upon by the descriptive analysis panelists and panel leader. After the panel evaluated attribute intensity of all the wines, ANOVA revealed that there were differences among the wines for each attribute. Blanc Du Bois wine quality was positively correlated with peach and rose and negatively correlated with greenwood/stemmy, phenolic /rubber bitter, and higher spectrophotometric absorbance at p < 0.10. P rincipal component analysis indicated that Blanc Du Bois wines tended to have one of two flavor profiles. Citrusy, bitter, and greenwood/stemmy wines tended to contrast with wines possessing sweet, fruity, and floral attributes. High quality was associated more with the latter group of attributes. The two Florida w ines in this study trended toward the citrus/woody side of the biplot and were grouped together by both the DA and volatile cluster analyses. The Louisiana wines did not group, and although some of the Texas wines did cluster together, there was no clear association with respect to flavor/aroma profiles among all the Texas wines on the PCA biplot. Certain volatiles in the wines correlated with sensory attributes. Ethyl and acetate esters in particular were often correlated with fruit and floral attributes. Isoamyl acetate (0.500) ethyl octanoate (0.482) ethyl decanoate ( 0.514) and ethyl dodecanoate (0.476) correlated positively with quality, and ethyl butanoate ( 0.485) ethyl 2 -
96 methylbutanoate ( 0.467) furfural ( 0.495) and linalool2 ( 0.514) correlated negatively with quality at p < 0.10. Gas chromatography olfactory was performed on two wine samples representing the two attribute categories identified by the DA PCA. Fifteen compounds 10 esters 3 alcohols, a terpene, and an organic acid were identified as odor active in the wine samples. Five of them all esters were shared between the two analyzed samples It is evident that there is substantial variation among Blanc Du Bois wines in terms of their flavor/aroma profiles and intensities, as well as their chemical markers and volatile profiles An examination of Blanc Du Bois viticultural and winemaking practices in the context of wine quality is the next logical investigatory step in the study of Blanc Du Bois wines. This research could identif y factors that influence the development or suppression of the volatile aroma compounds that are responsible for the desirable and undesirable sensory attributes identified in this study.
97 APPENDIX VOLATILE CONCENTRATI ONS Table A 1 Concentrations of volat iles detected by GC MS, in g/L. Odor active volatiles indicated by footnote Wine sulfur dioxide ethyl acetate ethyl isobutanoate1 isobutyl acetate ethyl butanoate1 n propanol ethyl 2methylbutanoate2 ethyl 3methylbutanoate hexanal isobutanol pinene iso amyl acetate1 butanol ethyl 2butenoate limonene isoamyl alcohol2 1 472.9 283.4 83.0 7.7 13.2 12.6 41.5 10.0 215.0 2.8 6.8 995.9 2 530.2 198.5 69.9 48.5 5.2 0.9 15.1 35.5 1517.1 7.6 814.8 3 376.5 77.6 87.4 5.9 10.0 41.8 1173. 7 1.9 1108.3 4 586.1 351.7 70.4 10.9 3.6 29.2 1152.9 860.8 5 370.6 81.5 7.9 19.4 1413.2 1.9 541.5 6 865.8 460.7 127.9 7.8 12.8 16.1 52.0 14.2 744.4 8.9 1114.7 7 866.2 341.1 104.0 7.2 8.8 17.2 125.4 599.9 0.3 2224.7 8 255.0 25.4 70.6 0.3 10.5 14.8 873.7 2.7 356.4 9 377.5 237.9 79.2 6.1 10.1 10.7 29.8 14.5 265.8 0.3 888.4 10 184.3 291.0 150.9 14.4 21.8 8.7 321.6 0.6 1.1 849.9 11 396.1 343.0 78.5 12.6 16.2 9.4 95.5 4.8 418.3 3.0 1192.5 12 871.2 318.6 87.0 10.2 23.3 13.8 48.2 20.0 164.4 963.9 13 13.3 382.1 242.2 116.0 9.1 18.1 11.3 26.9 7.0 356.0 0.8 0.8 869.7 14 482.9 264.3 45.0 150.2 7.3 4.4 13.7 54.2 2.4 845.1 3.0 3.2 10 32.2 1Confirmed on both wine sniffs. 2Confirmed only on wine 6 sniff s 3Confirmed only on wine 9 sniffs
98 Table A 1 Continued Wine ethyl hexanoate2 cis ocimene hexyl acetate p cymene terpinolene octanal 3 methyl 1 pentanol ethyl heptanoate3 ethyl lactate h exanol3 2 butoxyethanol methyl octanoate2 nonanal E 2 hexenol ethyl octanoate2 acetic acid3 1 1872.3 9.5 107.1 10.4 4.5 2.8 3.2 20.9 43.9 10.1 2.7 2.7 4286.3 39.3 2 2371.2 7 6.4 107.1 1.1 0.9 8.3 14.4 6.2 7401.9 25.7 3 2340.2 75.8 107.1 9.0 25.5 4.3 5892.9 8.9 4 1475.3 99.1 107.1 15.6 0.8 4330.6 34.6 5 1827.1 211.1 107.1 4.8 1.6 20.2 10.7 5525.4 9.0 6 2800.6 29.5 17 .5 107.1 18.1 4.7 5.3 113.6 10.9 7656.8 32.7 7 1515.3 107.1 6.3 6.8 1.7 170.3 7.3 1.7 1.0 3617.6 64.5 8 1314.4 147.7 107.1 1.7 8.6 18.8 3.9 3759.9 7.1 9 1311.2 5.3 107.1 6.9 8.8 45.8 5.7 3255.6 42.5 10 19 30.8 5.0 29.9 107.1 7.4 69.7 5.2 5251.9 27.8 11 1932.3 17.0 107.1 4.7 4.3 4.1 14.4 27.6 3.0 4567.4 17.2 12 1966.2 107.1 11.3 150.4 11.9 4829.7 72.2 13 1170.7 107.1 3.9 5.3 35.7 3.2 2.8 1.7 2880.5 24. 3 14 1607.5 3.4 107.1 5.1 1.1 2.9 125.1 7.7 1.7 3887.0 49.9 1Confirmed on both wine sniffs. 2Confirmed only on wine 6 sniffs. 3Confirmed only on wine 9 sniffs.
99 Table A 1 Continued Wine isoamyl hexanoate octyl acetate sabinene hydrate nerol o xide furfural decanal ethylsorbate ethyl nonanoate linalool2 linalool2 vitispirane benzaldehyde octanol methyl decanoate ethyl decanoate1 3 methylbutyl octanoate 1 2.9 14.4 0.8 15.2 300.6 6.7 6.2 48.0 2.6 4.7 1371.4 14.0 2 20.9 50.0 4.6 6.5 7.2 2791.6 19.4 3 26.0 5.5 293.7 12.1 5.1 9.8 4.3 6.1 1748.5 24.6 4 7.5 261.0 8.4 1.4 3.8 1319.3 5.4 5 11.7 0.4 5.0 245.7 6.8 14.6 11.0 1827.5 15.4 6 52.2 7.4 13.4 27.9 361.2 5.8 44.0 11.5 1.9 7.2 2518.4 20.5 7 15.2 0.8 5.6 653.3 21.6 8.6 13.9 5.6 773.2 19.5 8 8.0 2.8 3.0 6.7 7.2 1335.5 8.2 9 1.1 11.6 14.4 7.3 2.1 4.6 2.1 8.1 11.8 4.9 2.9 852.6 10.9 10 0.8 11.7 4.0 5.5 16.3 8.5 3.6 4.5 14 97.8 16.6 11 11.2 7.8 6.2 439.7 8.1 3.3 17.0 2.5 1089.3 15.5 12 3.3 19.9 7.8 14.6 14.4 4.3 68.0 2.9 5.8 1689.5 16.1 13 1.4 1.3 5.7 3.3 5.1 6.8 0.3 11.3 4.1 2.8 784.4 9.9 14 1.6 8.0 8.0 99.8 5.6 12.2 8.1 4.6 878 .4 11.0 1Confirmed on both wine sniffs. 2Confirmed only on wine 6 sniffs. 3Confirmed only on wine 9 sniffs.
100 Table A 1 Continued Wine butyrolactone ethylsuccinate ethyl9 decenoate terpineol valencene 8 heptadecene citronellol phenethyl acetate ethyl dodecanoate hexanoic acid phenethyl alcohol2 octanoic acid 1 123.5 14.8 10.8 17.7 26.7 102.1 56.5 2 23.9 12.5 1.4 20.1 68.9 40.4 91.5 105.5 3 5.5 26.2 3.9 44.0 17.8 83.8 46.3 4 4.3 16.0 22.3 10.9 20.7 31.1 56.5 73.7 5 6.6 9.9 16.4 66.4 37.7 58.0 98.6 6 79.9 16.8 11.1 15.1 55.1 54.4 69.3 119.5 7 300.0 132.0 6.7 9.8 22.6 181.5 39.7 8 2.6 7.0 5.7 0.6 29.4 16.7 20.9 50.8 9 4.1 55.5 5.2 10.4 10.4 24.5 76.2 59.4 10 16.9 23.8 28.8 46.7 77.6 11 10.2 53.4 23.8 45.2 24.0 80.0 52.5 12 9.8 112.3 16.3 17.6 22.7 65.5 54.5 13 1.7 36.5 27.4 11.3 8.2 21.9 48.1 44.4 14 58.3 2.8 3.0 4.6 16.6 15.7 62.7 41.1 1Confirmed on both wine sniff s. 2Confirmed only on wine 6 sniffs. 3Confirmed only on wine 9 sniffs.
101 REFERENCES Acree T, Arn H. 2004. Flavornet. Retrieved from http://flavornet.org/flavornet.html Date last a ccessed : June 2010. Ameri ne MA, Roessler EB. 1983. Wines: Their sensory ev aluation. Revised and e nlarged. San Francisco, CA: W.H. Freeman and Company. 432 p. Amerine MA, Ough CS, Ough CS. 1980. Methods for analysis of musts and wines. New York: Wiley. 341 p. Anderson M. 2006. Texas winemakers find success in awardwinning wine made from UFs hardy grape. Retrieved from http://news.ufl.edu/2006/07/12/uf grape/ Date last accessed: May 31, 2010. Barbe J, Pineau B, Ferreira. Antonio Cesar Silva. 2008. Instrumental and sensory a pproaches for the characterization of compounds r esponsible for w ine a roma. Chem Biodiv 5(6):117083. Bartowsky EJ, Henschke PA. 2004. The buttery attribute of winediacetyl desirability, spoi lage and beyond. Int J Food Microbiol 96(3):23552. Bitar A, Ghaddar T, Malek A, Haddad T, Toufeili I. 2007. Sensory t hresholds of Selected p henolic constituents from t h yme and their a ntioxidant p otential in s unflower o il. J Am Oil Chem Soc 85(7):6416 Bla ckman J, Saliba A. 2009. Sensory characterization of Hunter Valley Semillon using descriptive analysis. Flavor Fragrance J 24(5):23844. Carlucci A, Monteleone E. 2008. A p rocedure of sensory e valuation for d escribing the a roma p rofile of single g rape vari ety w ines. J Sensory Stud 23(6):81734. Charters S, Pettigrew S. 2007. The dimensions of wine quality. Food Qual Prefer 18(7):9971007. Civille GV, Lyon BG. 1996. Aroma and f lavor l exicon for sensory e valuation: t erms, d efinitions, r eferences, and e xamples 1st ed. West Conshohocken, PA: ASTM. 158 p. Damodaran S, Parkin K, Fennema O. 2007. Fennema's f ood chemistry. 4th ed. Boca Raton, FL: CRC Press. 1144 p. Eggers N. 2005. Varietal a roma c ompounds. Retrieved from https://people.ok.ubc.ca/neggers/Chem422A/VARIETAL%20AROMA%20COMPO UNDS.pdf Date last accessed: June 12, 2010.
102 p rofiling of f lavor a ttributes of w hite w ines from d ifferent g rape varieties. Int J Food Prop 10(3):6519 Fan W, Xu Y, Jiang W, Li J. 2010. Identification and q uantification of i mpact a roma compounds in 4 n onfloral Vitis vinifera varieties g rapes. J Food Sci 75(1):S818. Fontoin H, Saucier C, Teissedre P, Glories Y. 2008. Effect of pH, ethanol and acidity on astringency and bitterness of grape seed tannin oligomers in model wine solution. Food Qual Prefer 19(3):28691. GardeCerdn T, AncnAzpilicueta C. 2007. Effect of SO2 on the formation and evolution of volatile compounds in wines. Food Control 18(12):15016. Gurban A M Prieto Simn B, Marty J L Noguer T. 2006. Malate b iosensors for the m onitoring of m alolactic f ermentation: d ifferent a pproaches. Anal Lett 39(8):154358. Guth H. 1997. Quantitation and sensory studies of character i mpact o dorants of d ifferent w hite w ine v arieties. J Agric Food Chem 45(8):302732. Haak R. 2010. 2009 Blanc Du Bois p roduction & vineyard d ata. Blanc Du Bois acreage and production data assembled from vineyards Unpublished raw data. Herbert P, Cabrita MJ, Ratola N, Laureano O, Alves A. 2005. Free amino acids and biogenic amines in wines and musts from the Alentejo region. Evolution of amines during alcoholic fermentation and relationship with variety, subregion and vintage. J Food Eng 66(3):31522. Holzgang D. 1981. California wine list: A consumers guide to 144 Cabernets Los Angeles CA: Cheshire Bookseller s. 56 p. Hutkins RW. 2006. Microbiology and t echnology of f ermented f oods. 1st ed. Ames, IA : Blackwell Publishing. 473 p. Jones PR, Gawel R, Francis IL, Waters EJ. 2008. The influence of interactions between major white wine components on the aroma, flavour and texture of model white wine. Food Qual Prefer 19(6):596 607. King ES, Swiegers JH, Travis B, Francis IL, Bastian SEP, Pretorius IS. 2008. Coinoculated f ermentations u sing S accharomyces yeasts a ffect the volatile composition and sensory p roperties of Vitis vinifera L. cv. Sauvignon Blanc w ines. J Agric Food Chem 56(22):1082937. Klieber A, Bagnato N, Barrett R, Sedgley M. 2002. Effect of post ripening nitrogen atmosphere storage on banana shelf life, visual appearance and aroma. Postharvest Biol Technol 25(1):1524.
103 Komes D, Ulrich D, Lovric T. 2006. Characterization of odor active compounds in Croatian Rhine Riesling wine, subregion Zagorje. Eur Food Res Technol 222(1):1 7. Kotseridis Y, Baumes R. 2000. Identification of i mpact o dorants in Bordeaux r ed g rape j uice, in the commercial yeast u sed for i ts f ermentation, and in the p roduced w ine. J Agric Food Chem 48(2):4006 Labrouche F, Clark AC, Prenzler PD, Scollary GR. 2005. Isomeric influence on the oxidative coloration of phenolic compounds in a model white wine: comparison of (+) catechin and ( ) epicatechin. J Agric Food Chem 53(26):99938. Lawless H. 1995. Dimensions of sensory q uality: A critique. Food Qual Prefer 6(3):1919 Lawless HT, Heymann H. 1998. Sensory e valuation of f ood: p rinciples and p ractices. 1st ed. New York NY: Chapman & Hall. 827 p. Lawless H, Liu Y. 1997. Evaluation of w ine q uality u sing a smallp anel h edonic scaling m ethod. J Sensory Stud 12(4):31732. Lee S, Noble AC. 2003. Characterization of o dor a ctive compounds in Californi an Chardonnay w ines u sing GC Olfactometry and GC Mass Spectrometry. J Agric Food Chem 51(27):803644. Leffingwell and Associates. 1999. Odor & f lavor d etection t hresholds in w ater ( i n p arts per b illion). Retrieved from http://www.leffingwell.com/odorthre.htm Date last accessed: June 13, 2010. Lenoir J. 2006. Le Nez Du Vin. Z.I. 14 Avenue Gay Lussac 13470 Carnoux en Provence / France. Lund CM, Thompson MK, Benkwitz F, Wohler MW, Triggs CM, Gardner R, Heymann H, Nicolau L. 2009. New Zealand Sauvignon B lanc d istinct f lavor characteristics: sensory, chemical, and consumer a spects. Am J Enol Vitic 60(1):112. Maarse H. 1991. Volatile compounds in f oods and b everages. New York NY: CRC Press. 784 p. Mrtens son A. 2007. Need for protective measures to combat potential outbreaks of Homalodisca coagulata and Pierce's Disease in European viticulture. Acta Agr Scand B S P 57(1):19. Mateo JJ, Jimnez M. 2000. Monoterpenes in grape juice and wines. J. Chromatogr. A 881(12):55767.
104 Medley JC. 2003. Pierce's Disease. AgroEcoSystems Research Group, Texas A&M University. Retrieved from http://beaumont.tamu.edu/Research/Agroecosystems /Grapes/PierceDis.htm Date last accessed May 31, 2 010. Meilgaard MC, Civille GV, Carr BT. 2007. Sensory Evaluation Techniques. 4th ed. Boca Raton, FL: CRC Press. 448 p. Meilgaard MC. 1993. Individual differences in sensory threshold for aroma chemicals a dded to beer. Food Qual Prefer 4(3):15367. Mirarefi S, Menke SD, Lee SY. 2004. Sensory p rofiling of Chardonel w ine by d escriptive a nalysis. J Food Sci 69(6):S2117. Mortensen JA. 1987. Blanc Du Bois : a Florida bunch grape for white wine. Gainesville, FL: Agricultural Experiment Station, Institute of Food and Agricultural Sciences, University of Florida. 4 p. Noble AC, Arnold RA, Buechsenstein J, Leach EJ, Schmidt JO, Stern PM. 1987. Modification of a s tandardized system of w ine a roma t erminology. Am J Enol Vitic 38(2):1436 Noble AC, Strauss CR, Williams PJ, Wilson B. 1988. Contribution of t erpene g lycosides to b itterness in Muscat g rapes. Am J Enol Vitic 39(2):12931. Noble AC, Arnold RA, Masuda BM, Pecore SD, Schmidt JO, Stern PM. 1984. Progress toward a standardized system of wine aroma terminology. Am J Enol Vitic 35(2):1079 Noble AC, Shannon M. 1987. Profiling Zinfandel w ines by sensory and chemical a nalyses. Am J Enol Vitic 38(1):15. Nurgel C, Pickering G. 2005. Contribution of glycerol, ethanol an d sugar to the perception of viscosity and density elicited by model white wines. J Texture Stud 36(3):30323. Ough CS, Winton WA. 1976. An e valuation of the Davis Winescore Card and i ndividual e xpert p anel m embers. Am J Enol Vitic 27(3):13644. Perestrel o R, Fernandes A, Albuquerque FF, Marques JC, Cmara JS. 2006. Analytical characterization of the aroma of Tinta Negra Mole red wine: Identification of the main odorants compounds. Anal Chim Acta 563(12):15464. Prez Prieto LJ, Lpez Roca JM, Gmez Plaza E. 2003. Differences in major volatile compounds of red wines according to storage length and storage conditions. J Food Compos Anal 16(6):697705.
105 Pickering GJ, Heatherbell DA, Vanhanen LP, Barnes MF. 1998. The e ffect of e thanol co ncentration on the t emporal p erception of viscosity and d ensity in w hite w ine. Am J Enol Vitic 49(3):30618. Ramey DD, Ough CS. 1980. Volatile ester hydrolysis or formation during storage of model solutions and wines. J Agric Food Chem 28(5):92834. Rapp A. 1990. Natural flavour s of wine: correlation between instrumental analysis and sensory perception. Fresenius J Anal Chem 337(7):77785. Rapp A. 1998. Volatile flavour of wine: correlation between instrumental analysis and sensory perception. Nahrung / Food 42(06):35163. Reinec cius G. 2006. Flavor chemistry and t echnology. 2nd ed. Boca Raton, FL: CRC Press 489 p. Reynolds AG, Lowrey WD., Tomek L, Hakimi J, de Savigny C. 2007. Influence of i rrigation on vine p erformance, f ruit composition, and w ine q uality of Chardonnay in a cool, h umid climate. Am J Enol Vitic 58(2):21728. Rouseff R, Smoot J. 2006. Flavor d atabase. University of Florida Citrus Research and Education Center. Color and Flavor Chemistry Group. Retrieved from http://www.crec.ifas.ufl.edu/crec_websites/Rouseff/Website2002/Subpages/databa se_a_Frameset.htm Date last accessed: June 12 2010. Sauvageot F, Vivier P. 1997. Effects of m alolactic f ermentation on sensory p roperties of f our Burgundy w ines. Am J Enol Vitic 48(2):18792. Sefton MA, Francis IL, Williams PJ. 1993. The volatile c omposition of Chardonnay j uices: a study by f lavor p recursor a nalysis. Am J Enol Vitic 44(4):35990. Selli S, Canbas A, Varlet V, Kelebek H, Prost C, Serot T. 2008. Characterization of the m ost o dor a ctive volatiles of o range w ine m ade from a Turkish cv. Kozan ( Citrus sinensis L. Osbeck). J Agric Food Chem 56(1):22734. Sims CA, Mortensen J. 1989. Evaluation of w ine p otential from b reeding l ines and cultivars in Florida. Proc Fla State Hort Soc 102:21721 Skinkis PA, Bordelon BP, Butz EM. 2010. Effects of sunlight e xposure on b erry and w ine m onoterpenes and sensory characteristics of T raminette. Am J Enol Vitic 61(2):14756. Stone H Sidel JL. 2004. Sensory e valuation p ractices. 3rd ed. San Diego, CA: Elsevier Academic Press. 126 p. The Good Scents Company. 2010. F lavor materials. Retrieved from http://www.thegoodscentscompany.com/flavonly.html Date last accessed: June 15, 2010.
106 Torrens J, Riu Aumatell M, Vichi S, L pez Tamames E, Buxaderas S. 2010. Assessment of volatile and sensory p rofiles between b ase and sparkling w ines. J Agric Food Chem 58(4):245561. University of California Statewide Integrated Pest Management Program (UCIPM). 2008. UC m anagement g uidelines for Pierce's d isease on g rape. Retrieved from http://www.ipm.ucdavis.edu/PMG/r302 101211.html Date last accessed: May 31, 2010. van der Merwe CA, van Wyk CJ. 1981. The contribution of some f ermentation p roducts to the o dor of d ry w hite w ines. Am J Enol Vitic 32(1):416. Verzera A, Ziino M, Scacco A, Lanza CM, Mazzaglia A, Romeo V, Condurso C. 2008. Volatile compound and sensory a nalysis for the characterization of an Italian w hite w ine from Inzolia g ra pes. Food Anal Methods 1(2):14451. Vianna E, Ebeler SE. 2001. Monitoring e ster f ormation in g rape j uice f ermentations u sing solid p hase m icroextraction coupled with g as chromatographm ass spectrometry. J Agric Food Chem 49(2):58995. Vidal S, Francis L, Williams P, Kwiatkowski M, Gawel R, Cheynier V, Waters E. 2004. The mouthfeel properties of polysaccharides and anthocyanins in a wine like medium. Food Chem 85(4):51925. Vidal S, Williams P, Doco T, Moutounet M, Pellerin P. 2003. The polysaccharides of red wine: total fractionation and characterization. Carbohydr Polym 54(4):43947. Vilanova M, Sieiro C. 2006. Determination of free and bound terpene compounds in Albario wine. J Food Compos Anal 19(67):6947. Woods C. 2002. Blanc Du Bois: Floridas a ward w inning w ine w ins a gain! Retrieved from http://news.ifas.ufl.edu/2002/04/02/blanc dubois floridas award winning wine wins again/ Date last accessed: May 31, 2010. g lycerol by t wo e ndogenic w ine yeast strains at d ifferent i noculum size. Food Technol Biotechnol 44(4):5259 Yanniotis S, Kotseridis G, Orfanidou A, Petraki A. 2007. Effect of ethanol, dry extract and glycerol on t he viscosity of wine. J Food Eng 81(2):399403. Zea L, Moyano L, Moreno J, Cortes B, Medina M. 2001. Discrimination of the aroma fraction of Sherry wines obtained by oxidative and biological ageing. Food Chem 75(1):7984. Zoecklein BW, Fugelsang KC, Gump B H, Nury FS. 1995. Wine a nalysis and p roduction. 1st ed. New York, NY: Chapman & Hall. 621 p.
107 BIOGRAPHICAL SKETCH Eric Dreyer was born and raised in south Florida. He graduated cum laude with a Bachelor of Science in food science and human nutrition from t he University of Florida in May 2008. In August 2008 he began pursuit of a masters degree in food science with a minor in packaging science, also at the University of Florida. He has a strong interest in fermentation science and intends to pursue a career in that discipline.