Design and evaluation of a non-steady state rumen model

A dynamic simulation model of digestion and absorption of nutrients was modified and evaluated under non-steady state conditions. The results of detailed grazing experiments, including allowed grazing times (Exp. 1), combinations of rumen fill and starvation length before grazing (Exp. 2) and contrasting sward masses and sward heights (Exp. 3) as main treatments were used as reference values. The model was modified to run under a discontinuous feed input of ryegrass. Neutral detergent fibre (NDF) and nitrogen rumen pools were predicted with a relatively low root mean square prediction error (MSPE) of the observed means (11%) in Exps. 1 and 2, but a higher value (18%) was observed in Exp. 3. This significantly higher root MSPE was ascribed to the long period (up to 20.5 hours) of starvation that followed grazing. Prediction was poorer for organic matter rumen pool (root MSPE of 16%) than for NDF and N rumen pools, which requires further validation. Volatile fatty acid (VFA) rumen pool and VFA concentration were predicted with a root MSPE of the observed mean of 32-33%, which was close to the random variation observed in the experiments. Ammonia rumen pool was poorly predicted and the way ammonia is represented in the model must be modified to predict ammonia production and absorption under non-steady state conditions. The model can be used to predict ruminal digestion and absorption of nutrients (except ammonia) of grazing lactating dairy cows under discontinuous feeding regimens provided large periods of starvation are avoided.