Seasonal variation in Irish industry bovine milk composition characterised by chemical analysis and near‐infrared spectroscopy: A 12‐month study

Rationale Milk composition is an important factor affecting final cheese quality and yield, plant efficiencies, as well as determining the overall profitability of cheese processing plants. However, milk composition varies by season. Aim This study will investigate seasonal variation in key compositional components in Irish bovine milk and evaluate the potential of near‐infrared (NIR) spectroscopy for the rapid prediction of raw milk composition for informing upstream processing. Methods This study presents a comprehensive analysis of seasonal fluctuations in key milk components by examining 555 homogenised bovine milk samples from industry‐scale silos across Ireland over a 12‐month period. Simultaneously, the feasibility of near‐infrared (NIR) spectroscopy for rapid detection of key compositional components (i.e. total solids, total protein, fat, casein, lactose and ionic calcium) in raw milk prior to chymosin‐induced coagulation was evaluated. Principal component analysis (PCA) and partial least squares (PLS) regression were developed, with PLSR being assessed using coefficient of determination ( R 2 cv), root mean square error of cross‐validation (RMSEcv), ratio of standard error of prediction to sample standard deviation (RPD) and range of error ratio. Major Findings Wet chemistry analysis revealed clear seasonal trends in milk composition. PCA visualised distinct seasonal clustering, with winter samples clearly separated from spring and summer. High predictive accuracy was achieved for fat content ( R 2 cv = 0.92, RMSEcv = 0.11, RERcv = 14.53, RPDcv = 3.46), with good performance also observed for total protein and casein content prediction ( R 2 cv ≥ 0.81, RMSEcv ≤ 0.11, RERcv ≥ 9.85, RPDcv ≥ 2.31). However, predictive models for total solids, lactose and ionic calcium showed limited accuracy ( R 2 cv ≤ 0.59, RMSEcv ≥ 0.10, RERcv ≤ 7.17, RPDcv ≤ 1.54), indicating the need for further investigation. Industrial Implications Overall, this study bridges the gap between lab‐scale NIR method evaluation and industrial implementation by validating the approach using samples from large‐scale silos, marking an important step towards rapid, real‐time compositional analysis in dairy processing.