NIR spectroscopy to authenticate caprine cheeses from Grape-Pomace supplemented rations

The modern dairy industry, particularly the caprine sector, isthe focus of increasing scientific and commercial attention. Simultaneously,the drive towards a circular economy necessitates sustainable solutions foragricultural waste management. In this context, the use of grape pomace—anabundant by-product of the wine industry—as a feed supplement for livestockrepresents a promising strategy. However,introducing a dietary element like grape pomace into the feed regime of dairyanimals poses a fundamental analytical challenge: how to rapidly and reliablytrace the chemical footprint of such supplementation within a complex andcontinuously evolving matrix like cheese.The prima ry objective of this Master's thesis was to test thelimits and potential of Near-Infrared (NIR) Spectroscopy, coupled withchemometrics, for the authentication of dietary origin and the determination ofthe ripening stage in caprine cheeses.The experime ntal design involved Murciano-Granadina goats,located in Spain, fed with three different rations: a Control, and twosupplemented with Ensiled Red Grape Pomace (RGP) and Ensiled White Grape Pomace(WGP), respectively. The resulting milk was then used for cheese production,with samples analyzed at three ageing stages (days): 1, 30, 60 of ripening. The main objecti ve was to verify whether the subtle chemicalvariations induced by dietary supplementation were reliably detected by twodifferent NIR instruments: a portable (Neospectra) and a benchtop (FOSS-DS2500)and accurately classified using chemometric models.The methodological core lay in the dual nature of samplepreparation, which is a key focus of this thesis: whole cheese samples wereanalyzed for instrumental colour assessment. The spectroscopic NIR analysis and the relatedchemometric modeling were performed on grounded samples. Colorimetric analysis revealed that diet did not have astatistically significant main effect on cheese colour. However, a detailedanalysis highlighted a significant Ration x Ripening interaction for L* and a*parameters, with the RGP group distinguishing itself from CTR at Day 30,suggesting a dynamic modulation of oxidation induced by the grape pomace.Despite this, the Total Colour Difference (∆E) between cheeses supplementedwith black and white grape pomace was below the human visual perception threshold(∆E < 3).Classification analysis via N IR led to a primary scientificfinding: the optimal predictive model achieved a maximum Accuracy of only 54%for diet classification (NeoSpectra, full spectrum) and a maximum R²of 0.43 forripening time regression (NeoSpectra, full spectrum). These values,statistically close to random chance, demonstrate that the target chemicalvariation (diet-induced) is too subtle or masked by the high intrinsicheterogeneity of the caprine cheese matrix to be reliably resolved by the NIRtechnique.Despite the predictive failure, chem ometric optimizationprovided a crucial methodological contribution: the application of SpectralBand Selection enhanced the stability of the FOSS-DS2500 instrument, restoringthe expected hierarchy of instrumental performance and proving the need forintense normalization for low-resolution portable sensors (NeoSpectra). In conclusion, this research provides un equivocal evidence defining thelimitations of NIR spectroscopy for the direct authentication of caprinecheeses based on subtle dietary changes. The failure to achieve acceptableclassification accuracy and the highly inaccurate regression models forripening demonstrate that, despite rigorous sample preparation and advancedchemometrics, the necessary chemical signal was too dilute or overlaid to bereliably resolved by the NIR technique in this food matrix.