Evaluation of different energy feeding systems with production data from lactating dairy cows offered grass...

A set of data from 838 lactating dairy cows, drawn from 12 long-term feeding studies (at least 8 wk/period), was used to evaluate the energy feeding systems for dairy cows currently adopted in Australia, France, The Netherlands, the United Kingdom, and the United States. The animals were offered mixed diets of concentrates, forage [grass silages (n = 33) and corn silages (n = 5)] ad libitum. Data used in the present evaluation were either measured [dry matter (DM) intake, milk production and live weight], measured/estimated [dietary metabolizable energy (ME) concentration] or estimated [milk energy output and live weight change (LWC)]. The mean-square prediction error (MSPE) was used for the evaluation. Total ME intake, milk yields, and LWC varied from 91 to 338 MJ/d, 7.7 to 48.9, and -1.23 to 1.73 kg/d, respectively. Australian and French systems predicted total energy requirement and milk yield relatively well, while British, Dutch and American systems underpredicted total energy requirement by proportionately 0.06, 0.04, and 0.03, respectively; and overpredicted milk yield by 0.09, 0.06, and 0.04. The Agricultural and Food Research Council (AFRC) each produced a relatively larger error of the bias (predicted - actual data) over the total MSPE for ME requirement and milk yield and a relatively smaller error of random than other systems. However, an addition of proportionately 0.05 to the total predicted ME requirement of AFRC, as suggested in this system and currently used in the UK, indicated the prediction accuracy of ME requirement and milk yield is similar to Australian and French systems. Nevertheless, all the systems had a poor prediction of LWC. For each system, the total prediction error (total MSPE) was mainly derived from the line (slope; 0.49 to 0.64 of total MSPE), while less derived from the random (0.20 to 0.48 of total MSPE), indicating a large variation between the predicted and actual LWC existed among individual cows. The residual plots of the residual differences in LWC against predicted LWC revealed that the prediction error was greater with increasing LWC. It is concluded that Australian and French systems have a better prediction of total energy requirement and milk yield than other systems, and LWC is an inappropriate indicator of energy balance in lactating dairy cows.