Efficiency of different spectroscopies and the Electronic Nose techniques for the characterization of milk

Abstract

Four rapid methods, which are complementary to the usual MIR-based analyses, were compared in order to characterize local milk products. A set of 278 fresh samples from four separately reared Jersey, Piemontese and Valdostana cattle and Saanen goat herds was analyzed by: Fluorescence Spectroscopy, Electronic Nose, UV-Vis-NIRS and FT-NIRS (total 5851 digits by record). The Gross Composition and Fatty Acid composition were determined at the same time. Chemometric analysis of the digital measurements and of the milk composition was performed by discriminant PLS regression over the four herds. The average R2 cross-validated values of the six discriminant contrasts were lower for the Gross Composition (0.47), very high for the FT-NIRS scans (0.97), for the Fatty Acids (0.96), and also high for the Fluorescence (0.90) and the UV-Vis-NIRS evaluation (0.89), while the Electronic Nose gave lower distinction between the groups (0.64). The patterns based on the distance matrix showed a remarkable complementarity between the Gross Composition evaluation and the rapid methods, which were close to the Fatty Acids evaluation. The FT-NIRS and Fluorescence analyses converged together, clustering the Jersey & Piemontese, the Valdostana and then the Goat milk. The Jersey-Piemontese cluster was also confirmed by EN. The UV-Vis-NIRS appraisal, distinguished the Piemontese milk more clearly, while it paired the Jersey and Valdostana milk. These rapid methods could be of great interest in the milk research.