Modelling aroma of three Italian red wines by headspace-mass spectrometry and potential functions.

The aromas of 41 samples of wine from two Italian regions, Piedmont and Tuscany, were analysed by headspace-mass spectrometry. Samples were from three Italian wines (Barbera, Dolcetto and Chianti) produced in the same vintage, from different grape varieties and producing zones. The headspace generating conditions were optimised by full factorial experimental design then chemometric techniques were applied to verify the discriminating power of headspace-mass spectrometry among the three wine aromas. The modelling method based on potential function, applied on the first nine significant components of the 201 measured m/z, revealed best discrimination among the three wine aromas: cross-validated mean prediction rate of 96.7% and mean prediction rate of 83.3% on external test sets were obtained.