Evaluation of the dry matter intake predictions of the Cornell Net Carbohydrate and Protein System with Holstein and dual-purpose lactating cattle in the tropics

Data from three separate experiments were used to evaluate the dry matter intake (DMI) predictions of version 5.0 of the Cornell Net Carbohydrate and Protein System (CNCPS). Experiment 1 was conducted with 12 lactating Holstein cows confined in individual stalls, receiving known amounts of supplements and fresh, chopped Panicum maximum cv Tobiatá grass. Experiments 2 and 3 were conducted with 12 and 13 crossbreed dual-purpose cows rotationally grazing Cynodon nlemfuensis cv Alicia and P. maximum cv Tobiatá grass, respectively, with appropriate supplementation. Pasture intake in the grazing experiments was estimated for an 8-day period using the alkane technique. Model-predicted DMI based on measured feed analyses for the group of animals on Experiment 1 was very close to the observed values, with a mean bias of −0.19 kg DM per day, suggesting that the adjustment of DMI used by the CNCPS model in tropical conditions is accurate for confined lactating animals. The intake predictions by the CNCPS model for the grazing dual-purpose lactating cows were not as accurate. The CNCPS model underpredicted DMI in Experiment 2, and different degrees of accuracy and bias were evident in Experiment 3. For the three experiments, the slope of the regression between observed and predicted DMI was not different from unity, but the intercept was significantly different (P<0.05) from zero, suggesting a mean bias in the prediction. It should be recognized that accurate intake data on grazing animals is difficult to obtain and errors in the estimation of herbage intake using the alkane method could be contributing to the bias in the predictions by the CNCPS model.