Application of multivariate statistical analyses to the study of factors affecting white wine volatile composition

APPLICATION OF MULTIVARIATE STATISTICAL ANALYSES TO THE STUDY OF FACTORS AFFECTING WHITE WINE VOLATILE COMPOSITIONjfq_364 40..50 VALERIA GALLO1, RAFAEL BELTRÁN1, FRANCISCO J. HEREDIA2, M. LOURDES GONZÁLEZ-MIRET2and DOLORES HERNANZ1,3 1Department of Analytical Chemistry, Faculty of Experimental Sciences, University of Huelva, Avda. de las Fuerzas Armadas s/n 21007, Huelva, Spain 2Laboratory of Food Colour and Quality, Department of Nutrition and Food Science, Faculty of Pharmacy, University of Seville, Sevilla, Spain 3Corresponding author. TEL: 349-5921-9960; FAX: 349-5921-9942; EMAIL: [email protected] Received for Publication March 21, 2009 Accepted for Publication July 8, 2010 doi:10.1111/j.1745-4557.2010.00364.x ABSTRACT The influence of different storage conditions on the volatile composition of young white wine was evaluated during 1 year of storage. The wine was bottled and stored under different conditions of temperature, position and illumination. During storage, a significant increase in total content esters (mainly due to the increase in ethyl butyrate, ethyl acetate and diethyl succinate) and total content acids was detected. Also total content alcohols and carbonyl compounds decreased during storage. Storage time, temperature and illumination have a significant effect on esters content, whereas carbonyl compounds are affected by temperature, and alcohols content is affected by illumination. Stepwise linear discriminant analysis was applied to the experimental data.The results showed that the samples were well classified as a function of storage temperature. On the other hand, a certain trend was observed when the wine samples were classified according to storage time. The samples corresponding to two and four months of storage were well differentiated from the rest. PRACTICAL APPLICATIONS From a sensory point of view, the aroma influences greatly the quality of wines. The loss of aromatic characteristics in young white wine, as well as browning, take place over a variable period of time which ranges from one month to several years, depending on wine type and storage conditions. Assessment of possible factors affecting the useful lifespan of young white wine is of great interest for the wine making industry. Knowing the chemical composition of wine, specifically aromas, and its evolution relating to different storage and preservation conditions is an important tool in quality control. INTRODUCTION The loss of aromatic characteristics in young white wine, as well as browning, take place over a variable period of time, which ranges from one month to several years, depending on wine type and storage conditions. These chemical changes occurring during the wine storage are numerous and complex, and the process depends on grape factors, such as composition, pH or enzymatic activity (Ough and Ramey 1980; Singleton et al. 1985; Marais 1986; González-Viñas et al. 1996; Pérez-Prieto et al. 2003; Garde-Cerdán et al. 2008). From a sensory point of view, the aroma greatly influences the quality of wines. The aromatic composition of wines is very complex because there is a large number of volatile components in the wine (most of them being at very low concentrations), as well as to the interactions among them. Alcohols and esters are the major volatile components, although there Journal of Food Quality ISSN 0146-9428 40 Journal of Food Quality 34 (2011) 40–50 © 2011 Wiley Periodicals, Inc. are also variable amounts of aldehydes, acids, terpenes, volatile phenols and acetonic compounds. Aromas change over time; they evolve and become more complex, and wine becomes less fruity and gains bouquet due to acetalization and esterification reactions (Rapp et al. 1985). Acetals are products derived from the reaction of aldehydes with alcohols, with the most evident case being the formation of ethyl hemiacetal. Esterification reactions, which are very slow over time, produce highly characteristic compounds, such as ethyl acetate, which, when present at high concentrations (>140 mg/L), gives wine sour features (De Rosa 1998). The loss of fruitiness is often attributed to terpene oxidation reactions in wines with significant monoterpenol content (Ribereau-Gayon et al. 1975). During wine bottle storage, a number of changes affecting wine stability and organoleptic characteristics occur (Recamales et al. 2006). These changes will depend not only on wine chemical composition,but also on storage duration and conditions, such as temperature, oxygen availability or illumination. Pérez-Coello et al. (2003) studied the influence of storage temperature on ester hydrolysis and its relationship with the loss of fruitiness in wine.Additionally, there is a relationship between dimethyl sulphide levels and sensory evaluation of stored wines, depending on storage temperature (Marais and Pool 1980). In white wines stabilized without sulfurous anhydride, temperature during storage is the most significant factor in the variation of volatile composition (Garde-Cerdán et al. 2008). Storage time also affects aroma (González-Viñas et al. 1996). According to González-Viñas et al. (1998), sensory characteristics of young white wines stored under usual commercial conditions remain up to 18 months. Pérez-Prieto et al. (2003) studied the evolution of major volatile compounds present in red wines submitted to different periods of storage in bottles (8 and 10 months) and in oak barrels (7 months and 4 years). On the other hand, changes were observed during storage under controlled temperature (15.8C) and humidity (65.1%) for 5 years, depending on the type of bottle closure (cork or synthetic material) (Skouroumounis et al. 2005), the bottle position (horizontal and vertical position) during 36 months in white and red wines (Mas et al.2002), and the type of glass used for the bottle (Selli et al. 2002). Assessment of possible factors affecting the useful lifespan of young white wine is of great interest for the winemaking industry.Knowing the chemical composition of wine, specifically aromas, and its evolution relating to different storage and preservation conditions is an important tool in quality control. The aim of this study is to evaluate the effect of preservation environmental conditions (time, temperature, bottle position and illumination) on white wine quality characteristics. On the other hand, Stepwise Linear Discriminant Analysis (SLDA) was used to determine the most relevant changes in volatile composition in wine during storage. MATERIALS AND METHODS Samples The wine was produced in wineries adhered to the DO Condado de Huelva according to traditional methods, fermented at controlled temperatures. Healthy ripe grapes of Vitis vinifera cv. Zalema (white grape variety) grown exclusively in southwest Spain were destemmed, crushed and the must subsequently transferred into a 30,000-L stainless steel tank. Enough sulfur dioxide solution (3.1%) was added to adjust the total sulfur dioxide content of the must to 80 mg/L. Themustwaskeptinthetankat12 ⫾3Cfor24 h,andthenthe residualsolidswereseparated.Lateron,themustwasrunintoa fermentation tank, and the total acidity was adjusted to 6 g/L with tartaric acid. For the fermentation processes,no selected yeasts were used and the temperature was kept at 22 ⫾2C.At the end of this stage, the wine was transferred to a 14,000-L stainless steel tank and it was stored for the stabilization process.Theproductwaskeptat12 ⫾3Cduringthestabiliza- tion process, and the free sulfur dioxide content was maintained at 25–30 mg/L. After cold stabilization, the wines were bottledin750-mLgreenglassbottlesandstoredundertwodif- ferent conditions of temperature: “variable,” where bottles were stored under variable conditions of temperature (daily and seasonal, within the approximate range from 8 to 30C), and “constant,” where temperature ranged between 15 and 20C throughout the year.Within each of these lots (“variable” and “constant”), half of the bottles were placed in horizontal positionandtheotherhalfinverticalposition;bothamongthe horizontal bottles and the vertical ones; half of the samples were stored in the dark and the other half under variable light conditions (daylight and night). The wine was stored for 12 months. The samples were taken at the moment of bottling, andthensamplingfrequencywasdonebimonthly.Atotalof98 samplesof bottled winewereanalyzed. Reagents and Standards All reagents used were of chromatographic quality: dichloromethane, ammonium sulfate and tartaric acid (Merck, Darmstadt, Germany), ethanol (Panreac Química S.A.U., Barcelona, Spain) and potassium bitartrate (Sigma-Aldrich, St. Louis, MO). The standards used were acetaldehyde, 2-methylpropanoic acid and 3-methylbutanoic acid (Dr Ehresnstorfer GmbH, Augsburg, Germany); acetoin, butanoic acid, hexanoic acid, ethyl isovalerate, octanoic acid, isoamyl acetate, ethyl lactate, 2-phenylethanol and decanoic acid (Aldrich Chemical); furan-2-carbaldehyde,ethyl acetate, 2-methylpropyl ethanoate, 2-phenylethyl acetate, ethyl butyrate, ethyl octanoate, diethyl succinate, butan-1-ol, isobutanol, isoamyl alcohol and hexan-1-ol (Chem Service Inc); ethyl hexanoate and cis-3-hexen-1-ol (Fluka Chemika). V. GALLO ET AL.STATISTICAL STUDY OF WHITE WINE VOLATILES 41Journal of Food Quality 34 (2011) 40–50 © 2011 Wiley Periodicals, Inc. The internal standards used were 4-methyl-2-pentanol (Chem Service) and 2-octanol (Aldrich). All standards had a purity rate between 98 and 99%. The deionized water used was provided by the Milli-Q filtration system (Millipore model A-10). The standard solutions were prepared using ethanol as solvent at a concentration of 1,000 mg/L. The concentrations of each internal standards compounds in the solution (just one solution for both standards) were 743 mg/L of 4-Methyl- 2-pentanol and 620 mg/L of 2-octanol. The wine model solution was prepared by weighing 1.16 g potassium bitartrate and 0.7 g tartaric acid dissolved in 1,000 mL of ultrapure water with 10% absolute ethanol (pH =3.10) (Zhou et al.1996). To achieve synthetic wine,the wine model solution (250 mL) was strengthened with a solution of aromas,so that concentration of the 25 compounds in the synthetic wine was 50 mg/L. The synthetic wine was used for optimization of the chromatographic method and for the precision study of the extraction method. The solutions were stored at 4C until use. Analysis of Volatile Compounds Preparation of the sample was carried out following the methods proposed by Ortega et al. (2001). 9 mL of sample were placed in a conical-bottom centrifuge tube, adding 13.5 g of ammonium sulfate, 21 mL of deionized water and 100 mL of the internal standard solution (final concentration 2.5 mg/L of 4-methyl-2-pentanol and 2.1 mg/L of 2-octanol). The mixture was shaken for 30 min, adding afterwards600 mLdichlorometane,furthershakenfor 30 min,and centrifuged at 3,000 rpm for 10 min. The organic phase was collected and 1 mL was injected into the chromatograph for analysis. All analyses were performed in triplicate and the organic extract was injected in duplicate into the gas chromatograph. The result was expressed as mean value. The volatile compounds were analyzed by gas chromatography using the chromatographic system Agilent Technologies 6,890 N Network GC System (Palo Alto, CA) equipped with a flame ionization detector and a Varian CP-Wax 52 CB column (60 m ¥0.25 mm ¥0.25 mm) (Walnut Creek, CA). The injection (1 mL of sample) was performed using the split mode (split ratio 1:20) and oven temperature was held constant at 40C for 5 min, then raised by 1.5C/min. to 200C, where it was held for 20 min. The injector and detector temperature was 250C. The carrier gas used was helium at a flow rate of 1 mL/min. Analytical Parameters of the Extraction Method Quantification of volatile compounds was performed using the internal standard method. From standard solutions, calibration lines were constructed at six different concentration levels, taken into account the usual concentrations of these analytes in the wines (0.5 to 150 mg/L for esters and alcohols and 0.1 to 50 mg/L for acids and carbonyl compounds). The regression coefficients (r2) ranged between 0.9830 (for butanoic acid) and 0.9998 (for decanoic acid). The values of repeatability (precision) were assessed by calculating the relative standard deviation (R.S.D.) in six independent extractions from the synthetic wine, performed under the same analytical conditions as in the sample in the shortest time possible (UNE 82009–1:1998). Values ranged between 1.6% R.S.D for 2-phenylethyl acetate and 12.4% R.S.D. for hexanoic acid. Multivariate Statistical Analysis The various multivariate statistical treatments were performed using the STATISTICA® statistical package (version 6.0, StatSoft 2001). Variance analysis was applied to the data obtained for the variables studied,whichallowsus tocomparethe meanvalues obtained for different groups or classes, thus establishing whether there are significant statistical differences between them among the wines stored under different conditions, or whether these differences are due to random error. On the other hand, SLDA was applied for the characterization and differentiation of the samples according to their aromatic composition. RESULTS AND DISCUSSION Twenty-five volatile compounds were identified in the wine samples,which could be classified into four different families: (1) esters (ethyl isovalerate, isoamyl acetate, ethyl lactate, ethyl acetate, 2-methylpropyl ethanoate, 2-phenylethyl acetate, ethyl butyrate, ethyl octanoate, diethyl succinate and ethyl hexanoate), (2) acids (2-methylpropanoic acid, 3-methylbutanoic acid, butanoic acid, hexanoic acid, octanoic acid and decanoic acid), (3) alcohols (2- phenylethanol, butan-1ol, isobutanol, isoamyl alcohol, hexan-1-ol and cis-3-hexen-1-ol); and (4) carbonyl compounds (acetaldehyde, acetoin and furan-2-carbaldehyde). Tables 1 to 3 show the effect of the different storage conditions (time, temperature, illumination and bottle position) on the volatile contents (means and standard deviations) of the wines. Ethyl acetate,butanoic acid and acetoin concentrations in the wine are not affected significantly by any storage factor. Considering the sum of all individual volatile compounds, which could mean the global aroma, the factors “storage time” and “temperature” have significant effect on the concentration of volatile compounds (P<0.01 and P<0.001, STATISTICAL STUDY OF WHITE WINE VOLATILES V. GALLO ET AL. 42 Journal of Food Quality 34 (2011) 40–50 © 2011 Wiley Periodicals, Inc. TABLE 1. INFLUENCE OF TIME AND VARIOUS STORAGE CONDITIONS ON ESTERS (mg/L)† Variable Ethyl lactate Ethyl hexanoate Ethyl octanoate Ethyl butyrate Ethyl isovalerate Diethyl succinate 2-Phenyl ethyl acetate Isoamyl acetate 2-methylpropyl ethanoate Ethyl acetate Storage time (months) 0 25.41 ⫾4.58and 0.04 ⫾0.01a0.02 ⫾0.01a0.01 ⫾0.00a6.70 ⫾2.24a0.11 ⫾0.05a0.40 ⫾0.13a0.46 ⫾0.22a15.51 ⫾8.15a 2 21.31 ⫾2.81a0.03 ⫾0.02a0.04 ⫾0.01a0.03 ⫾0.00a0.01 ⫾0.00a4.78 ⫾1.24b0.09 ⫾0.06a0.84 ⫾0.29b0.27 ⫾0.11a17.35 ⫾8.56a 4 25.28 ⫾2.96a0.03 ⫾0.01a0.05 ⫾0.02a0.04 ⫾0.01b0.01 ⫾0.00b5.69 ⫾1.11b0.08 ⫾0.04a0.70 ⫾0.14b0.38 ⫾0.26a18.67 ⫾7.04a 6 26.13 ⫾279ab 0.03 ⫾0.01a0.04 ⫾0.01a0.03 ⫾0.01a0.01 ⫾0.00b10.16 ⫾3.44c0.09 ⫾0.05a0.52 ⫾0.15c0.31 ⫾0.27a25.04 ⫾9.59a 8 31.55 ⫾2.95b0.02 ⫾0.02a0.05 ⫾0.01a0.03 ⫾0.01a0.01 ⫾0.00b13.66 ⫾4.69d 0.08 ⫾0.06a0.29 ⫾0.32d 0.38 ⫾0.26a23.17 ⫾10.07a 10 26.82 ⫾3.41ab 0.03 ⫾0.01a0.05 ⫾0.01a0.04 ⫾0.01cb 0.01 ⫾0.00c14.52 ⫾3.35d 0.08 ⫾0.04a0.35 ⫾0.23d 0.37 ⫾0.21a23.90 ⫾8.28a 12 25.19 ⫾3.23a0.02 ⫾0.02a0.05 ⫾0.02a0.03 ⫾0.01a0.01 ⫾0.00c15.68 ⫾3.77d 0.08 ⫾0.04a0.25 ⫾0.18d 0.32 ⫾0.23a22.96 ⫾10.59a Significant effect *** ns ns * ns *** ns *** ns ns Temperature Variable 26.52 ⫾4.08 0.02 ⫾0.01 0.05 ⫾0.01 0.03 ⫾0.01 0.01 ⫾0.00 13.02 ⫾5.86 0.08 ⫾0.06 0.31 ⫾0.28 0.50 ⫾0.18 23.21 ⫾10.44 Constant 25.58 ⫾5.09 0.04 ⫾0.01 0.04 ⫾0.01 0.02 ⫾0.01 0.01 ⫾0.00 8.47 ⫾3.41 0.08 ⫾0.03 0.68 ⫾0.24 0.18 ⫾0.12 20.49 ⫾7.64 Significant effect ns *** *** *** ns *** ns *** *** ns Illumination Light 27.50 ⫾4.55 0.03 ⫾0.02 0.04 ⫾0.01 0.03 ⫾0.01 0.01 ⫾0.00 11.15 ⫾4.32 0.08 ⫾0.05 0.50 ⫾0.26 0.35 ⫾0.22 23.66 ⫾9.69 Darkness 24.59 ⫾4.22 0.03 ⫾0.01 0.05 ⫾0.02 0.03 ⫾0.01 0.02 ⫾0.00 10.34 ⫾4.24 0.09 ⫾0.05 0.48 ⫾0.27 0.33 ⫾0.21 20.03 ⫾8.45 Significant effect ** ns ns ns * ns ns ns ns ns Position Horizontal 26.15 ⫾5.05 0.03 ⫾0.02 0.04 ⫾0.01 0.03 ⫾0.01 0.01 ⫾0.00 10.85 ⫾5.62 0.10 ⫾0.05 0.48 ⫾0.32 0.41 ⫾0.26 19.92 ⫾8.32 Vertical 25.95 ⫾4.14 0.03 ⫾0.01 0.05 ⫾0.01 0.03 ⫾0.00 0.01 ⫾0.00 10.64 ⫾4.93 0.07 ⫾0.04 0.51 ⫾0.31 0.27 ⫾0.15 23.77 ⫾9.71 Significant effect ns ns * ns ns ns * ns *** ns Significant differences at * P<0.05; ** P<0.01; *** P<0.001. † Data expressed as mean ⫾SD. Note: Data followed by different letters for each column are significantly different according to Fisher′s test, P<0.05. ns, not significant; nd, not detected. V. GALLO ET AL.STATISTICAL STUDY OF WHITE WINE VOLATILES 43Journal of Food Quality 34 (2011) 40–50 © 2011 Wiley Periodicals, Inc. respectively), whereas bottle position and illumination have no significant effect on the variation of volatiles during storage (data not shown). The variance analysis of the individual volatile compound data shows that temperature during storage (variable/ constant) is a factor that significantly affects a greater number of volatiles, specifically 16 compounds (6 esters, 4 acids, 4 alcohols and 2 carbonyl compounds). Bottle position (horizontal/vertical)has a significant effect on11 compounds (3 esters, 3 acids, 4 alcohols and furan-2-carbaldehyde); illumination (light/darkness) has a significant effect on 5 compounds (2 esters, 2 acids and isoamyl alcohol), and storage time on 5 compounds (isobutanol and 4 esters). The total content of esters and acids (that is, considering the sum of all individual esters and acids compounds,respectively) increases during storage, whereas alcohol and carbonyl contents (considering the sum of all individual alcohol and carbonyl contents, respectively) decrease. The increase in ester content in the wines is mainly due to the increase in ethyl butyrate, ethyl acetate and diethyl succinate. Thus, after 12-month storage, wines have 1.5 times more ethyl butyrate and ethyl acetate, and 2.3 times more diethyl succinate. Ethyl lactate and isovalerate concentrations were the only component significantly modified by light intensity. Samples stored inthe darkhad lowerconcentrationof ethyl lactate thanthose exposed to light, whereas the wines stored in the dark had higher concentration of ethyl isovalerate than the wines exposed to light. Regarding 2-phenylethyl acetate, isoamyl acetate and 2-methylpropyl ethanoate concentrations, they decrease during storage. In the case of isoamyl acetate, the decrease is much higher in the wines stored at variable temperature, while in the case of 2-methylpropyl ethanoate, the decrease is higher in the wines stored at constant temperature (Table 1). Studies carried out in wines elaborated with the Airen,Viura and Macabeo white grape varieties show similar results in the evolution of esters during bottle storage (González-Viñas et al. 1996; Pérez-Coello et al. 1999, 2003). Storage time does not affect acid content significantly, whereas temperature during storage significantly affects the content of 2-methylpropanoic, 3-methylbutanoic, octanoic anddecanoicacids,with thewines storedat variable temperature showing higher concentration of 2-methylpropanoic, 3-methylbutanoic and decanoic acids (Table 2). Exposition to light affects decanoic and hexanoic acid concentration significantly, while bottle position affects the content of 2-methylpropanoic, 3-methylbutanoic and decanoic acids in the wines. With respect to alcohols, temperature is the factor that most affects its concentration, with the wines stored at variable temperature and horizontal position showing higher TABLE 2. INFLUENCE OF TIME AND VARIOUS STORAGE CONDITIONS ON ACIDS (mg/L)† Variable 2-methylpropanoic acid 3-Methylbutanoic acid Butanoic acid Hexanoic acid Octanoic acid Decanoic acid Storage time (months) 0 nd 0.44 ⫾0.20a16.96 ⫾4.28a5.58 ⫾0.34a2.17 ⫾0.99a1.54 ⫾0.35a 2 nd 0.15 ⫾0.23a10.92 ⫾6.07a5.08 ⫾0.44a3.09 ⫾1.10a1.51 ⫾0.20a 4 0.24 ⫾0.16a0.22 ⫾0.21a11.89 ⫾4.81a5.30 ⫾0.19a3.50 ⫾0.61a1.58 ⫾0.09a 6 0.11 ⫾0.10a0.19 ⫾0.15a11.43 ⫾3.27a5.00 ⫾0.40a2.71 ⫾0.95a1.58 ⫾0.39a 8 0.28 ⫾0.27a0.25 ⫾0.19a11.50 ⫾4.74a5.16 ⫾0.44a3.02 ⫾1.10a1.73 ⫾0.45a 10 0.24 ⫾0.25a0.27 ⫾0.23a12.63 ⫾3.74a5.10 ⫾0.24a3.37 ⫾1.26a1.47 ⫾0.10a 12 0.20 ⫾0.17a0.16 ⫾0.18a12.90 ⫾3.42a5.14 ⫾0.30a2.83 ⫾0.87a1.34 ⫾0.58a Significant effect ns ns ns ns ns ns Temperature Variable 0.31 ⫾0.29 0.29 ⫾0.21 12.99 ⫾2.49 5.11 ⫾0.31 2.55 ⫾0.70 1.68 ⫾0.33 Constant 0.05 ⫾0.10 0.13 ⫾0.09 10.76 ⫾5.36 5.15 ⫾0.38 3.63 ⫾0.94 1.39 ⫾0.32 Significant effect * ** ns ns *** *** Illumination Light 0.22 ⫾0.17 0.22 ⫾0.19 12.50 ⫾3.27 5.24 ⫾0.28 3.26 ⫾0.91 1.62 ⫾0.32 Darkness 0.14 ⫾0.26 0.20 ⫾0.18 11.25 ⫾5.10 5.02 ⫾0.37 2.92 ⫾1.06 1.44 ⫾0.37 Significant effect ns ns ns * ns * Position Horizontal 0.31 ⫾0.29 0.26 ⫾0.23 12.14 ⫾3.87 5.18 ⫾0.29 3.00 ⫾0.93 1.67 ⫾0.34 Vertical 0.05 ⫾0.14 0.15 ⫾0.14 11.61 ⫾4.73 5.08 ⫾0.39 3.17 ⫾1.05 1.40 ⫾0.32 Significant effect * * ns ns ns ** Significant differences at * P<0.05; ** P<0.01; *** P<0.001. † Data expressed as mean ⫾SD. Note: Data followed by different letters for each column are significantly different according to Fisher’s test, P<0.05. ns, not significant. nd, not detected. STATISTICAL STUDY OF WHITE WINE VOLATILES V. GALLO ET AL. 44 Journal of Food Quality 34 (2011) 40–50 © 2011 Wiley Periodicals, Inc. TABLE 3. INFLUENCE OF TIME AND VARIOUS STORAGE CONDITIONS ON ALCOHOLS AND CARBONYL COMPOUNDS (mg/L)† Variable hexan-1-ol 2-phenylethanol Isobutanol Butan-1-ol Isoamyl alcohol cis-3-Hexen-1-ol Acetaldehyde Acetoin Furan-2-carbaldehyde Storage time (months) 0 1.83 ⫾0.35a123.07 ⫾13.38a1.56 ⫾0.82a0.71 ⫾0.25a123.81 ⫾32.57a0.38 ⫾0.07a95.08 ⫾16.50a4.72 ⫾2.20a2.87 ⫾0.86a 2 1.10 ⫾0.42a75.67 ⫾20.73a0.52 ⫾0.52a0.70 ⫾0.13a125.59 ⫾17.68a0.31 ⫾0.07a91.05 ⫾11.36a4.84 ⫾2.07a1.47 ⫾0.47a 4 1.01 ⫾0.14a67.02 ⫾9.95a1.51 ⫾0.97a0.93 ⫾0.29a100.29 ⫾62.42a0.25 ⫾0.10a90.24 ⫾25.50a4.32 ⫾3.50a1.79 ⫾0.61a 6 1.07 ⫾0.31ab 69.95 ⫾25.55a1.02 ⫾0.92a0.70 ⫾0.27a120.97 ⫾10.48a0.32 ⫾0.09b83.99 ⫾16.36a5.27 ⫾1.70ab 1.86 ⫾1.02a 8 1.22 ⫾0.28ab 75.46 ⫾27.80a1.22 ⫾0.98a0.78 ⫾0.22a129.14 ⫾11.88a0.32 ⫾0.05b92.34 ⫾14.78a5.90 ⫾1.54ab 2.29 ⫾1.22a 10 1.02 ⫾0.28a62.52 ⫾19.10a1.38 ⫾0.99a0.80 ⫾0.37a121.94 ⫾9.73a0.29 ⫾0.03ab 83.29 ⫾15.71a7.12 ⫾1.76b2.00 ⫾0.74a 12 1.37 ⫾0.48b71.71 ⫾19.45a1.20 ⫾0.92a0.68 ⫾0.18a103.34 ⫾43.37a0.30 ⫾0.03ab 78.81 ⫾13.40a5.86 ⫾1.64ab 1.89 ⫾0.91a Significant effect ns ns * ns ns ns ns ns ns Temperature Variable 1.32 ⫾0.36 83.94 ⫾23.59 2.15 ⫾0.96 0.79 ⫾0.32 109.31 ⫾34.21 0.33 ⫾0.06 79.28 ⫾12.62 5.53 ⫾2.62 2.24 ⫾1.01 Constant 0.94 ⫾0.21 56.83 ⫾11.04 0.14 ⫾0.12 0.74 ⫾0.17 124.44 ⫾29.66 0.27 ⫾0.06 93.96 ⫾17.23 5.57 ⫾1.71 1.52 ⫾0.41 Significant effect *** *** *** ns ns *** ** ns ** Illumination Light 1.14 ⫾0.32 72.52 ⫾22.26 1.09 ⫾1.00 0.76 ⫾0.18 128.31 ⫾13.51 0.29 ⫾0.07 86.78 ⫾19.39 5.78 ⫾1.73 1.79 ⫾0.78 Darkness 1.12 ⫾0.39 68.26 ⫾24.55 1.20 ⫾1.06 0.77 ⫾0.31 105.45 ⫾41.82 0.31 ⫾0.07 86.46 ⫾13.25 5.32 ⫾2.62 1.98 ⫾0.94 Significant effect ns ns ns ns * ns ns ns ns Position Horizontal 1.20 ⫾0.38 78.70 ⫾28.22 1.42 ⫾1.04 0.87 ⫾0.28 118.09 ⫾26.77 0.32 ⫾0.07 85.71 ⫾11.41 5.76 ⫾2.36 2.12 ⫾1.12 Vertical 1.06 ⫾0.32 62.08 ⫾11.37 0.86 ⫾0.81 0.66 ⫾0.17 115.67 ⫾38.24 0.28 ⫾0.07 87.53 ⫾20.77 5.35 ⫾2.05 1.65 ⫾0.32 Significant effect ns *** ** ** ns ** ns ns * Significant differences at * P<0.05; ** P<0.01; *** P<0.001. † Data expressed as mean ⫾SD. Note: Data followed by different letters for each column are significantly different according to Fisher′s test, P<0.05. ns, not significant. V. GALLO ET AL.STATISTICAL STUDY OF WHITE WINE VOLATILES 45Journal of Food Quality 34 (2011) 40–50 © 2011 Wiley Periodicals, Inc. contents in the studied alcohols. On the other hand, storage time and exposition to light have no significant effect on carbonyl compounds. These results agree with studies carried out by other authors in white wines at various storage temperatures (commercialization and 5C) (Marais 1986; Pérez- Coello et al. 2003). Multivariate Statistical Analysis The existence of statistical significant differences in the levels of volatiles compounds as a function of the conditions of storage surveyed, were evaluated by one-way analysis of variance (ANOVA) and two-wayANOVA The latter method is an extension to the former that evaluates the effect of two independent variables concurrently, that is, two-way ANOVA is a procedure that enables to assess whether the two independent variables interact with respect to their effect on the dependent variable. The Fvalues and significance levels of the factors as well as their interaction on the concentration of volatile compounds are shown in Table 4.The results show that esters have significant differences (P<0.05) among the wines stored under the testedconditions(time,temperatureandillumination during storage).Ontheotherhand,acidsconcentrationisnotaffected bythedifferentstorageconditionstested.Illuminationexertsa significant influence on alcohol concentration, and during storage,temperaturesignificantlyconditionsthecomposition of carbonyl compounds in the wine. When two-way ANOVA was applied, three significant interactions appeared: time/ illumination for acids, and temperature/position and temperature/position/illuminationfor carbonylcompounds. The aim of SLDA is to study to what extent it is possible to divide two or more sample groups. By applying SLDA it is possible to establish functions that discriminate between previously established groups, in order to apply them later to originally non-classified cases (Castino 1975; Schreier and Reiner 1979; Gorostiza et al. 1982; González-San José and Diez 1993).One of the most commonly used methods within SLDA is the iterative procedure, in which variables are added one by one to the discriminant functions, yielding a study of discrimination after each step. The process concludes when no discrimination significantly better than the last is obtained. The SLDA was performed to determine which variables are the most appropriate to discriminate among samples. As a result, four SLDA were performed taking into account the different storage conditions, time, temperature, illumination and bottle position. Taking storage time into account, the mathematical model selects 11 variables (diethyl succinate, isobutanol, ethyl lactate, 2-phenylethanol, cis-3-hexen-1-ol, hexanoic acid, 2-methylpropyl ethanoate, ethyl octanoate, isoamyl alcohol, butan-1-ol and hexan-1-ol) with the greatest discriminatory power. The coefficients of the functions that classify the samples according to storage time are gathered in Table 5. By applying these functions, their validity is checked according to the success rate of the cases in their corresponding group. The classification matrix is shown in Table 6, with a mean classification rate of 89.6%.When the samples are graphically represented within the discriminant space (Fig. 1), it can be observed that the considered variables divide the wine samples by storage time. As observed in the chart, discriminant function 1 separates the samples corresponding to 2 and 4 months of storage from the rest. Discriminant function 1 is mainly related with diethyl succinate, cis-3 hexenol and hexan-1-ol with positive sign, and with isobutanol, 2-phenylethanol, ethyl octanoate and isoamyl alcohol with TABLE 4. MULTIVARIATE ANALYSIS OF VARIANCE BY STORAGE CONDITIONS Factors Esters Alcohols Acids Carbonyl compounds Fratio Pvalue Fratio Pvalue Fratio Pvalue Fratio Pvalue Time (T) 7.26 0.000* 1.24 0.307 0.31 0.905 0.74 0.600 Temperature (Ta) 7.79 0.008* 1.90 0.175 1.68 0.202 8.63 0.005* Bottle position (P) 1.27 0.265 3.62 0.064 0.63 0.431 0.04 0.844 Illumination (I) 6.36 0.015* 6.60 0.014* 2.22 0.144 0.01 0.904 T¥Ta 0.67 0.647 2.12 0.096 0.26 0.927 1.46 0.237 T¥P 1.77 0.156 1.15 0.360 0.47 0.793 0.85 0.522 T¥I 0.72 0.610 1.14 0.357 3.60 0.009* 0.28 0.916 Ta ¥P 0.08 0.772 2.29 0.143 4.14 0.053 28.72 0.000* Ta ¥I 2.02 0.161 0.00 0.933 0.03 0.856 0.14 0.707 P¥I 1.72 0.195 2.34 0.132 3.03 0.088 1.31 0.257 T¥Ta ¥P 0.22 0.949 0.307 0.903 0.19 0.961 0.38 0.854 T¥Ta ¥I 1.41 0.254 2.03 0.109 0.52 0.753 0.70 0.624 T¥P¥I 0.81 0.549 0.21 0.958 0.78 0.569 0.13 0.982 Ta ¥P¥I 0.01 0.906 0.07 0.785 2.81 0.10 5.86 0.020* * Significant effects at P<0.05. Note: Fratio and Pvalues obtained by Fisher′s test (P<0.05). STATISTICAL STUDY OF WHITE WINE VOLATILES V. GALLO ET AL. 46 Journal of Food Quality 34 (2011) 40–50 © 2011 Wiley Periodicals, Inc. negative sign. These volatiles compounds (isobutanol, 2-phenylethanol, ethyl octanoate and isoamyl alcohol) are impact odorants of Zalema wines (Gómez-Míguez et al. 2007) and are associated with fruit, floral, burned and fusel attributes, respectively (Campo et al. 2006). On the other hand, discriminant function 2 is related with 2-methylpropyl ethanoate with positive sign and ethyl lactate, hexanoic acid and butanol with negative sign. Taking into account the temperature during the storage period, the mathematical model selects 9 variables having the greatest discriminatory power with a classification rate of 100% in all cases. The canonical function is related with isobutanol, 2-phenylethanol, ethyl octanoate and ethyl acetate with negative sign, and butan-1-ol, furan-2- carbaldehyde, hexanoic acid, acetaldehyde and diethyl succinate with positive sign. In the graphical representation of the samples within the discriminant space (Fig. 2), it can be observed that the variables used divide the samples into two clearly distinguishable groups, one corresponding to the wines stored at variable temperature and another to the wines TABLE 6. CLASSIFICATION MATRIX % correct Storage time (months) 24681012 2 100.0 80000 0 4 87.5 17000 0 6 87.5 00701 0 8 100.0 00080 0 10 75.0 00016 1 12 87.5 00001 7 Total 89.6 97798 8 TABLE 5. COEFFICIENTS OF THE CLASSIFICATION FUNCTIONS (SLDA) ACCORDING TO STORAGE TIME Storage time (months) 24681012 Diethyl succinate -3.047 -3.879 -0.975 -0.301 0.246 1.070 Isobutanol 0.132 1.261 -2.895 -5.601 -5.039 -6.262 Ethyl lactate -0.734 -0.266 0.243 0.981 0.460 0.371 2-phenyl ethanol -0.005 -0.118 -0.306 -0.447 -0.449 -0.505 Cis-3-Hexen-1-ol -12.797 -46.956 22.334 26.091 22.579 29.438 Hexanoic acid 58.679 65.842 53.504 52.831 53.949 54.911 2-methylpropyl ethanoate -8.261 -26.626 -18.164 -27.759 -22.002 -21.686 Ethyl octanoate 820.605 838.479 563.093 602.710 582.633 408.273 Isoamyl alcohol 0.094 0.068 0.039 0.030 0.019 -0.037 Butan-1-ol -5.609 1.332 1.338 9.387 7.503 3.991 Hexan-1-ol 2.935 2.652 9.219 13.289 11.101 17.791 Constant -152.592 -168.946 -139.656 -160.938 -154.134 -156.672 root 1 root 2 2 months 4 months 6 months 8 months 10 months 12 months -6 -4 -2 0 2 4 6 -4 -3 -2 -1 0 1 2 3 4 FIG. 1. SCATTERPLOT OF SAMPLES IN THE PLANE DEFINED BY THE CANONICAL FUNCTIONS WHEN THE TIME OF STORAGE IS CONSIDERED FOR DISCRIMINATION V. GALLO ET AL.STATISTICAL STUDY OF WHITE WINE VOLATILES 47Journal of Food Quality 34 (2011) 40–50 © 2011 Wiley Periodicals, Inc. stored at constant temperature. The changes in the concentration of esters could be correlated with the loss of freshness aromas during the storage time and these sensory and chemical changes are less pronounced in wines stored at 5C (Pérez- Coello et al. 2003). Taking into account the bottle position during the storage period, the variables with the greatest discriminatory power were decanoic acid, four alcohols (butan-1-ol, cis-3-hexen-1- ol, isobutanol and 2-phenylethanol) and five esters (isoamyl acetate, ethyl octanoate, 2-methylpropyl ethanoate, ethyl butyrate and ethyl isovalerate). The calculated discriminant function allows distinction between the wines stored in horizontal position and those stored in vertical position, as observed in Fig. 3. Finally, the discriminant analysis applied root 1 samples variable Temperature constant Temperature -8 -6 -4 -2 0 2 4 6 0 10 20 30 40 50 60 FIG. 2. DISPERSION CHART OF WINES CLASSIFIED BY STORAGE TEMPERATURE USING THE CALCULATED DISCRIMINANT FUNCTION root 1 samples horizontal position vertical position -5 -4 -3 -2 -1 0 1 2 3 4 0 10 20 30 40 50 60 FIG. 3. SCATTERPLOT OF SAMPLES IN THE PLANE DEFINED BY THE CANONICAL FUNCTIONS WHEN THE BOTTLE POSITION IS CONSIDERED FOR DISCRIMINATION STATISTICAL STUDY OF WHITE WINE VOLATILES V. GALLO ET AL. 48 Journal of Food Quality 34 (2011) 40–50 © 2011 Wiley Periodicals, Inc.