Main content area

Evaluating model predictions of partitioning nitrogen excretion using the dairy cow model, Molly

Johnson, H.A., Baldwin, R.L.
Animal feed science and technology 2008 v.143 no.1-4 pp. 104-126
mathematical models, ruminant nutrition, model validation, prediction, dairy cows, nutrient partitioning, nitrogen, excretion, dynamic models, nitrogen balance, digestion
Computer models of dairy cow digestion and metabolism are useful tools to evaluate rations and estimate nitrogen (N) excretion. Using models, different management strategies and/or feeding practices can be evaluated and their impact on N excretion explored before implementation. Critical parameters for developing nutrient management plans can be identified to help reduce the impact of animal manure. However, before a model is implemented, errors associated with predictions and limitations of the model must be known. Otherwise, proposed management and feeding changes may not result in effective measures to optimally manage N excretion. The objective of this research effort was to determine whether or not Molly, a dynamic model of dairy cow digestion and metabolism [Baldwin, R.L., 1995. Modeling Ruminant Digestion and Metabolism. Chapman & Hall, UK], could be used to predict N excretion. In Molly, predictions of N excretion are based on level of milk production (ucells) and rate of insoluble protein hydrolysis (KPiAa). Accurate estimation of N excretion patterns would indicate that understanding of N metabolism as represented in Molly is adequate. Failures indicate that more research is needed. To examine estimates of N excretion, N balance data from 18 publications in the Journals of Dairy and Animal Science were simulated using Molly. Input data required included diet composition (N intake), initial bodyweight, days in milk, dry matter intake and milk production. A total of 73 different simulations (73 different diets) with N intakes ranging from 0.19 to 0.78kg/d and crude proteins of 106-206g/kg were run. Model outputs of milk, milk protein, faecal N, urinary N and milk N were compared to observed data. Overall, the model simulated reality well. Mean bias are less than 11% for milk yield, milk protein, faecal N, urinary N and milk N with coefficients of determination of 0.94, 0.73, 0.70, 0.80 and 0.88, respectively. Mean absolute errors and root mean square prediction errors are also low compared with observed mean milk, milk protein, faecal N, urinary N and milk N. N excretion losses due to volatilisation were not included in model estimates of faecal N and urinary N. Predicted rates of insoluble protein hydrolysis (KPiAa) from estimating N excretion indicated that current estimates of fractions of undegraded protein for individual feedstuffs in Rumen Nitrogen Usage [National Research Council (NRC), 1985. Ruminant Nitrogen Usage. National Academy Press, Washington, DC] are not adequate.