STOCHASTIC-MODELS OF GRAZING BEHAVIOR IN SHEEP

Three models, the multinomial, the Markov process and the age-dependent process, were used to simulate the behaviour of grazing sheep (Ovis sp.). These simulations were compared with data for sheep grazing perennial ryegrass (Lolium perenne L.) or white clover (Trifolium repens L.) swards. The Markov process was found to give an adequate representation of the bout structure and was examined further algebraically. It was shown that errors in estimation of grazing time for different sampling intervals using this model followed a similar pattern to that seen in the data. However, owing to lack of independence among the animals the Markov process overestimated the error whilst the multinomial model was shown to underestimate the error.