Potential of visible-near infrared spectroscopy for the characterization of butter properties

This study is aimed to investigate the potential of visible-near infrared (VIS-NIR) spectroscopy for the determination of butter characteristics. Analyzes (butter yield, texture, color, dry matter, water and fat content) were performed on 88 butters prepared in duplicate from 44 individual cow milks. In order to maximize the variability in the measurements, milk samples were collected on basis of several criteria, e.g. breeds and moment of sampling (morning or evening milking). Butter analyzes were carried out the day of the production. Butter samples were then frozen and the VIS-NIR spectra were carried out later after thawing in a refrigerator at 13 °C. Standard normal variate transformation followed by a first derivative treatment were applied on the spectra. The principal component analysis of the VIS-NIR spectra allowed to distinguish butters made from different breeds (i.e. Jersey vs Holstein). Using partial least square regression (PLS) and VIS-NIR spectra with leave-one-out cross-validation (cv), high determination coefficients (R2 cand R2cv) were obtained for the water content and color. For the water content, R2 c was 0.91 and R2cv was 0.88. For the color, R2c and R2 cv were respectively 0.93 and 0.91. These results indicated the possibility of using VIS-NIR spectral analysis for the simultaneous determination of mater content. The use of VIS-NIR spectra could be a tool to determine directly characteristics of new butter samples. Additional samples should be analyzed in order to consolidate the database and to validate with an independent dataset the models constructed.