Efficacy of near infrared spectroscopy to segregate raw milk from individual cows between herds for product innovation and traceability

Abstract

Cows with specialised characteristics and requirements can be aggregated into different herds for targeted nutritional management and to facilitate on-farm segregation of raw milk for the production of high-value niche dairy products, offering improved economic returns. Rapid methods for independent verification of product quality and origin are desirable to support validation and traceability of such products. This study examined the use of near infrared spectroscopy (NIRS) to segregate raw milk from individual cows of multiple breeds from different herds fed on the same or differing feeding regimes, and to correlate and evaluate the efficacy of the predictions for crude protein and the milk fatty acid (FA) phenotypes for each of the herds. Reference values and near infrared spectra were obtained from representative freeze-dried raw milk samples (n = 220) collected from 847 lactating cows of 3 breeds from the Lincoln University dairy farm in New Zealand. The feed sources (i.e. pasture or pasture with lucerne silage) significantly influenced the protein and the FA values, and these differences were reflected in NIRS analyses. The partial least square regression models for crude protein determination showed excellent results, whereas for the most dominant FA, they were not appreciable. Maximum separation was obtained between the herds on the same feeding regime (mean specificity = 95.2%) using the partial least square discriminant analysis, and its overall performance in differentiating the objects was better than that of the soft independent modelling of class analogy. The multiclass analyses conducted in this study offer improvements to current approaches for evaluating and validating raw milk for the manufacture of specific dairy products, and for enhancing product traceability.