CHEMOMETRIC CHARACTERIZATION AND CLASSIFICATION OF 5 VENETIAN WHITE WINES

A chemometric multiyear study of the characterization and classificati on of wines by chemical variables is presented. Five typical Venetian white wines (from the Veneto and Friuli regions of Italy) are consider ed: Soave Classico, Prosecco di Conegliano-Valdobbiadene, Verduzzo del Piave, Tocai di Lison, and Tocai delle Grave del Friuli. Data collect ed concern 273 samples originating from vintages between 1977 and 1984 , and 27 chemical analytical variables (18 inorganic and classical det erminations and 9 aroma compounds). The year-to-year variability is di scussed. The multivariate statistical methods include the classificati on by Euclidean distances from group centroids, the K-nearest-neighbor rule, principal component analysis (and its derived method SIMCA), an d canonical variate analysis. Statistical analysis reveals clear group structures in the data set and an acceptable wine separation in the m ultidimensional space. Best results are obtained by canonical variate analysis (overall correct classification about 90%), which shows that several chemical variables contribute to the classification, cis-3-hex en-1-ol being by far the most important, followed by 1-hexanol, potass ium, nitrogen compounds, and total phenols.