Development and evaluation of a model to predict herd-level milk fat concentration in Irish spring-calving grazing dairy herds.

The objective of this quantitative study was to mathematically reparametrize a herd-level milk fat concentration model, based on grazing rotation, into a continuous day-of-year model. An additional objective was to evaluate the predictive performance of the reparameterized model using milk recording data from 5,165 spring-calving (Feb. to Apr.) herds in Ireland. The dataset included 13,788 individual herd-level milk sample recordings collected between February and November 2023. The grazing-rotation coefficients from the original model were aligned to their corresponding day-of-year, and continuous daily coefficients were generated using linear extrapolation. The daily coefficients were then integrated into the original multivariable linear regression model, replacing grazing rotation with day-of-year. The reparameterized model was initially evaluated for prediction accuracy and precision using the evaluation dataset reported by Heffernan et al. (2025a), and it demonstrated strong predictive ability (R2 = 0.78; root mean square error [RMSE] = 0.24%; concordance correlation coefficient [CCC] = 0.85). Model prediction accuracy and precision were further assessed using a national dataset of 5,165 spring-calving herds with intermittent milk recording records from February to November 2023. The model also demonstrated strong predictive ability for the spring-calving dataset (R2 = 0.65; RMSE = 0.32%; CCC = 0.76). These results demonstrate that the reparameterized day-of-year model accurately predicts herd-level milk fat concentration across the grazing season (Feb. to Nov.) for Irish spring-calving dairy herds. The model provides a practical tool for distinguishing expected seasonal variation in milk fat concentration from other mechanisms that reduce milk fat synthesis, thereby supporting informed management decisions and enhancing production efficiency.