| Abstract: | Wines from three important wine-producing regions, Stellenbosch, Robertson, and Swartland, in the Western Cape Province of South Africa, were analyzed by ICP-MS and the elemental composition used in multivariate statistical analysis to classify the wines according to geographical origin. The method is based on the assumption that the provenance soil is an important contributor to the trace element composition of a wine. A total of 40 elements were determined in 40 wines. Of these, 20 elements: Li, B, Mg, Al, Si, Cl, Sc, Mn, Ni, Ga, Se, Rb, Sr, Nb, Cs, Ba, La, W, Tl, and U showed differences in their means across the three areas. In a stepwise discriminant analysis procedure, functions based on linear combinations of the log-transformed element concentrations of Al, Mn, Rb, Ba, W, and Tl were generated to correctly classify wines from each region. In an alternative approach, a pairwise discriminant analysis procedure, not previously used in wine provenance studies, was tested. In this procedure, the classification was done in three steps, with each step classifying a wine as coming from a certain region or not. The combination of elements characterizing wines from a particular region was different in each region. The discriminant functions were based on the following elements: Al, Mn, Rb, Ba, and W for Stellenbosch; Se, Rb, Cs, and Tl for Robertson; and Al, Mn, Rb, Sr, Ba, and Tl for Swartland. After this procedure, the classification of the wines into one of the groups was 100% successful. |
