The specificities ranged between 44% and 100%: the best results were obtained with the first derivative of the NIR spectra. In general, the first and second derivatives allowed to build QDA-UNEQ models with 100% sensibility and high specificity.
Subsequently, the potential of the fusion of information from the three instruments was investigated: the variables selected by STEP-LDA from the 14 matrices studied were joined and a new matrix with 105 variables was built. Then, STEP-LDA was repeated on this new matrix in order to select a lower number of relevant variables; six variables were retained: 4 NIR, 1 UV–vis and 1 Nose.
The QDA-UNEQ models, both for Chianti Classico and Liguria Riviera dei Fiori, built using these 6 fused variables were very ef-ficient: cross validation prediction abilities, sensitivities and spe-cificities 100%.