This paper reports a study carried out to determine the effect of post-filter-drying moisture uptake on the repeatability of NIR spectroscopy analyses of ewe milk using the dry-extract system for infrared (DESIR). The study assessed the spectral repeatability of the method and the effect of moisture uptake on that repeatability, using graphical representations and root mean squared (RMS) error between spectra, as well as multivariate Mahalanobis distances (GH values) and predicted constituent values (protein, fat and dry matter). The spectral repeatability of this method of dried milk analysis using glass-fibre filters was very high, both spectrally per se [RMS MEAN = 3.914 µlog (1/R) and STD = 4.126 µlog (1/R)] and in terms of NIR spectroscopy prediction of chemical components and of the Mahalanobis distance. Using a single filter per milk sample is thus considered sufficient to guarantee the required accuracy and precision of the DESIR method. Differences between spectra due to the presence or absence of air exposure [STD = 6446 µlog (1/R)] were greater than the method repeatability error. Predictions of protein, fat and dry matter were significantly higher for air-exposed filters, due to moisture uptake; the increase in GH values for air-exposed filters was highly significant, suggesting that milk samples to be analysed using the DESIR method should be stored in a desiccator after drying and before NIR analysis and not left exposed to air.

Keywords: near infrared spectroscopy, NIR, DESIR, desiccator, repeatability, ewe milk, chemical composition, water absorption