|De Souza, Rodrigo - Michigan State University|
|Tempelman, Robert - Michigan State University|
|Spurlock, Diane - Iowa State University|
|Armentano, Lou - University Of Wisconsin|
|Allen, Mike - Michigan State University|
|Vandehaar, Michael - Michigan State University|
Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 2/28/2017
Publication Date: 6/24/2017
Citation: De Souza, R., Tempelman, R.J., Spurlock, D.M., Armentano, L.E., Connor, E.E., Allen, M.S., Vandehaar, M.J. 2017. Development of equations to predict dry matter intake of lactating cows using animal factors. Journal of Dairy Science. 100(Suppl. 2):358 (abstr. 328).
Technical Abstract: Our objective was to model dry matter intake (DMI, kg) in Holstein dairy cows based on milk energy (MilkE, Mcal/d), energy required for maintenance, change in body weight (DeltaBW, kg/d), body condition score (BCS, scale 1 to 5), height (Htcm, cm), and parity. The database contained weekly DMI of 4,031 lactations from 3,393 Holstein cows from research stations across the US. The averages and standard deviation for the covariates were 24.9 ± 4.3 kg DMI, 29.8 ± 6.1 Mcal/d MilkE, 125 ± 12 kg BW^0.75, 626 ± 80 kg BW, 124 ± 12 kg BWBCS3^0.75, 620 ± 82 kg BWBCS3, 0.36 ± 1.29 kg/d DeltaBW, 3.00 ± 0.43 BCS, and 149 ± 6 cm Htcm, where BWBCS3 is the BW adjusted for BCS 3. Four full models were generated to model DMI wherein each model contained 1 of the 4 ways used to express BW, as showed above. The full models contained the fixed effects of the covariates described previous, parity, and all possible 2-way interactions between parity and the other covariates. Cow, diet, experiment, and location were included as random effects. The full models were first subjected to forward selection. The resulting models were then analyzed using HP MIXED from SAS 9.4, where the non-significant covariates (P > 0.05) were removed. Because the covariate parity was highly significant but the contrast between primiparous and multiparous was not significant, the weighted effect of parity based on the number of observations was added to the intercept. The final models were compared based on the root mean square error of prediction (RMSEP), decomposition MSEP, mean bias, and concordance correlation coefficient (CCC). The selected model used was BW expressed as kilograms: DMI = 2.58 + 0.30 * MilkE + 0.027 * BW + 0.050 * DeltaBW -1.15 * BCS (RMSEP = 2.61 kg, Mean bias, %MSEP = 0.61, Slope bias, %MSEP = 0.10, Mean bias = -0.20 kg, and CCC = 0.77). The selected model has smaller bias than the equations suggested by NRC (2001, mean bias, %MSEP = 32.61, slope bias, %MSEP = 4.61) to predict DMI and has a potential to benefit nutritionists during diet formulation.