Investigating Daily Changes in Food Intake by Ruminants

摘要

Trajectories of food intake for 12 sets of animals (weekly average intakes of four individual growing sheep, mean daily intakes of four pens of three growing cattle, and the mean daily intakes of four groups of six pregnant fallow does) were examined.The raw data exhibited expected seasonal trends, including changes in the amount of food eaten with changing live weight (LW) and age. The data were detrended by fitting polynomial equations of time. The residuals between the predicted and actual valueswere examined to determine the nature of period-to-period variations in food intake. Although the data-sets were obtained in different ways, the detrended data all showed similar food intake behaviours. In all but one case the residuals distributions were skewed negatively. The residuals vs time plots showed that large deviations below the expected values were more likely than deviations above the expected intakes, and that similar types of deviation tended to be clustered. Autoregressive integrated moving average (ARIMA) modelling indicated that autoregressive (AR) and/or moving average (MA) models, usually with a seasonal component, best fitted these data. However, the models had limited ability to predict food intake trajectories in the long term.These results suggest that when models are used to predict day-to-day variations in food intake, they should be primed by measuring actual intakes over at least 7-10 days; that similar intake behaviours are likely to occur together but that intakes thatare lower than predicted will occur more often than those that are greater than predicted; and that intakes will additionally fluctuate in an apparently random way.