MODELS OF VOLUNTARY FOOD-INTAKE IN CATTLE

Voluntary food intake in cattle has a great impact on performance. Consequently, much attention has been focused on prediction of intake and many models have been proposed. The objective of this paper is to discuss the context of different models, their advantages and disadvantages for feed management on farms and in feed evaluation. For dairy cows and growing cattle models were divided into simple or multiple regression models, more complex systems and fill systems and then tabulated. Dynamic models are discussed. Regression models are the most abundant whilst more complex models are scarce. Fill-systems have only been put to use in France and Denmark. Only a few dynamic models have been developed. Factors relating to animals, food, management, housing and environment influence intake regulation and are therefore important in predicting voluntary intake. How and which factors that have been included in predicting intake are discussed, i.e. animal factors such as breed, sex, live weight, age, parity, stage of lactation, milk yield, pregnancy, body condition, nutritional status and diseases; food factors such as diet and chemical composition, digestibility, energy concentration, degradation profiles, rate of passage, physical form, conservation, silage dry content and fermentation quality, palatability, fill-values; management factors such as time of access to feed, frequency of feeding, separate feeding versus complete diets, use of anabolic agents, food additives, mineral salts or alkaline agents; housing factors such as tie stalls versus loose housing, space allowance and space at manger; environmental factors such as photoperiod and temperature. Shortcomings and advantages in existing models for predicting intake and future research requirements are discussed. The following conclusions are made: Models should at least include important animal and food factors. Models should consider if cows are primiparous or multiparous. None of the models are able to accurately estimate intake in early lactation. Models assuming constant intake per kg W0.75 may result in serious lack of fit. Food, management and housing factors are inadequate in most models. Regression models often cover limited areas of feeding and are difficult to improve unless new experimental data can be added to existing data. More complex systems like the French and Danish fill-systems are more general and have the important advantage of being easily modified when new information becomes available. None of the existing models are able to consider interactions between nutrients from different feeds in the diet, and to do so dynamic and mechanistic models are needed.