Predicting milk fat concentration from nutrient content and DCAD of the diet
The farm-gate price of milk in Canada is based on composition, which provides incentive for producers to increase fat content. The objective of this study is to determine the extent to which changes in nutrients and DCAD can predict milk fat percentage. Data recorded by Valacta (Dairy Production Center of Expertise Quebec-Atlantic) for the years 2009 to 2011 was used and originally comprised 3,481,705 test-day records (275,758 cows and 3,140 herds). Records used for the regression analysis were restricted to those from Holstein cows, between 1 and 305 DIM, taken during winter months, reducing the number of admissible records for the analysis to 306,191 (134,236 cows and 2,658 herds). Lactations were divided into early (1-50 DIM), peak (51-100 DIM), and established lactation (101-305 DIM). Statistical analysis were performed using Proc HPMIXED of SAS with herd and cow(herd) as random effects. Independent variables were included in the final equation when P≤0.05. The variables used as covariates in the regression were: milk production (kg/day), DIM and estimated breeding value for fat composition (EBV_FAT). Variables tested to explain milk fat concentration were: NDF from forage + 0.5 x NDF from concentrate (NDF_NRC), NFC, amount of buffers (BUFF; kg/day), amount of fat supplements with more than 80% of palmitic acid (PALM80; kg/day) and DCAD. In the final analysis for the three years, multiple regression in early lactation (n=24,987; R2=0.44) included, in addition to the covariates, the following variables: NDF_NRC (quadratic) and PALM80. For peak lactation (n=29,317; R2=0.42) the variables were DCAD, NDF_NRC, BUFF and PALM80. For established lactation (n=100,706; R2=0.64) NDF_NRC, NFC (quadratic), BUFF and PALM80. All the equations accounted for a significant effect of year. When the regression was split by year, all the variables remained the same, while the DCAD (quadratic) was also added to the model for established lactation in all three years, but with different optimal value (2009: >330 mEq/kg MS; 2010: 210 mEq/kg MS; 2011: >380 mEq/kg). In summary, the equations were able to predict up to 64% of milk fat variation based on the combination of different nutrients, especially NDF_NRC (quadratic) and PALM80. Based on the variations in optimal values between years, it could be concluded that the impact of the DCAD on milk fat concentration can be influenced by the nutritional quality of feed ingredients.
Keywords: Lactating dairy cows, milk fat, DCAD.