Chemometric classification of some European wines using pyrolysis mass spectrometry

The application of Curie-point pyrolysis mass spectrometry (Py/MS) to 33 certified wine samples from different regions of Spain, France and Italy is reported. The resulting mass spectra were submitted to principal component analysis, canonical variates analysis and cluster analysis producing a final dendrogram, using average linkage clustering. A good classification was achieved according to large geographical areas (country of origin), while no correlation was found to grape species or to local areas of cultivation. The same data were submitted to artificial neural networks analysis using the back-propagation algorithm and dividing the data into a training and a test data set. More than 90% of the test wine samples were correctly identified according to their country of origin after training of the artificial neural network. Again no correlation was found between pyrolysis mass spectra and grape species or area of cultivation. Nevertheless, the technique could be applied to rapid screenings of wine samples for the tentative identification of country of origin. In combination with other techniques (IR-MS, NMR) it can give complementary information about the samples and their regions of origin.