Nutritional and non-nutritional factors associated with bulk tank milk fat concentration in Irish commercial spring-calving grazing dairy herds

The objectives of this experiment were (1) to identify associations between nutritional and non-nutritional factors and bulk tank milk fat concentration and (2) to develop a multivariable model capable of predicting herd-level milk fat concentration for Irish commercial spring-calving grazing dairy herds. An observational experiment comprising 25 commercial spring-calving dairy herds was conducted over a 2-yr period. Farms were visited 10 times per year, which coincided with each grazing rotation. During each visit, grassland measurements and pasture samples were collected from the next 2 paddocks to be grazed. Concentrate, silage, and other supplementary ingredients were sampled if included as part of the diet at each visit. Bulk tank milk samples were collected, along with data on pasture management, herd management, and herd genetic characteristics. Using the data from 12 of the 25 herds (i.e., model development data set), univariate analysis was performed to identify the relationships between each explanatory variable and milk fat concentration. Variables with a univariate analysis of P < 0.2 were included in a multivariable linear regression model, and backward stepwise elimination was performed until the remaining variables had P < 0.05 and the most parsimonious model was achieved. The predictive performance of the multivariable linear regression model was evaluated using the data from the remaining 13 herds (i.e., model evaluation data set). Across the whole data set, average bulk tank milk fat concentration was 4.54% ± 0.50%, with a range of 3.56% to 6.09%. The lowest average milk fat concentrations were observed during grazing rotations 3 and 4, coinciding with the late spring and early summer periods (i.e., May to early June), with the highest average milk fat concentration observed during grazing rotation 10. The average concentrations of de novo, mixed, and preformed milk fatty acids were 27.9 ± 2.0, 32.5 ± 1.6, and 37.5 ± 3.3 g/100 g of fat, respectively. Although several univariate relationships were identified, backward stepwise elimination identified a multivariable linear regression model with grazing rotation and the herd's milk fat concentration PTA as factors associated with bulk tank milk fat concentration. For the model evaluation data set, an initial multivariable linear regression model predicted bulk tank milk fat concentration with an R² of 0.73 and a root mean square error (RMSE) of 0.23%. The slope between observed and predicted bulk tank milk fat concentration was 1.15 (SE = 0.06), with an average bias of 0.05% and a relative prediction error (RPE) of 1.09. When a final model was evaluated, which was developed for all 10 rotations, the model predicted bulk tank milk fat concentration with an R² of 0.79 and a RMSE of 0.23%. The slope between observed and predicted bulk tank milk fat concentrations was 1.09 (SE = 0.04) with an average bias of 0.05% and an RPE of 1.06. Although the association between grazing rotation and bulk tank milk fat concentration is likely multifactorial, involving many nutritional and non-nutritional factors, the positive relationship between milk fat concentration PTA and milk fat concentration highlights the important role of genetics on milk fat production in pasture-based systems.