Predictions of food intake in ruminants from analyses of food composition

Equations used to predict intake by cattle from some chemical or physical characteristic of food were examined. The equations are empirical or mechanistic in nature. Mechanistic equations are not used widely, usually only in a research context. The input to mechanistic models requires too much time to quantify to be used routinely. Empirical relationships form the basis of most feeding standards and are based on a wide variety of prescribed characteristics (digestibility, chemical composition, etc.), but the underlying principle is a relationship between intake and digestibility. Equations are modified to take account of feed types, animal weight and physiological state, rumen modifiers, hormone implants, environmental conditions, and whether grazing or hand fed. Quite significant differences exist between the equations in the intakes they predict in response to variation in weight, breed type, and feed digestibility. Equations can be very precise in their prediction when used with feed types and breed types on which they are based. Near infrared reflectance (NIR) offers the most potential for long-term development of equations. At present, NIR is used largely to determine chemical composition because of speed of operation, but long-term storage of data is simple, allowing further associative relationships to be developed readily. More sophisticated statistical procedures being employed to improve the precision of the relationships between intake and prescribed characteristics of food and NIR will be vitally important as they enable extra parameters to be incorporated at no extra cost or time for analysis.