Impact of Variation in Ration Nutrient Inputs on Animal Responses Predicted by Two Dairy Cattle Metabolic Models

The objectives were to assess impacts of variation in ration nutrient levels (i.e., inputs) on metabolic model-predicted animal responses (i.e., outputs) to determine if variations in ration nutrient values are important to model-predicted dairy cattle outputs. The models used were Molly and CPM Dairy. The ration nutrient inputs varied were soluble carbohydrate (SC), starch, CP, soluble CP (SCP), NDF, ADF, lignin, ether extract, Lys, and Met. Three contemporary California commercial rations for 'high group' lactating dairy cows were used and nutrients were varied individually, +20% or −20% from base values. Animal outputs used to assess impacts of changes in ration nutrient inputs on model predictions were fecal N, urinary N, milk N (Molly only), microbial DM, milk (ME, metabolizable protein, and amino acid (AA) allowable milk in CPM Dairy), milk fat (Molly only), and metabolizable protein allowable milk protein in CPM Dairy. Results were evaluated by comparing relative sensitivities, which are the changes in model output in response to changes in nutrient input normalized for the average nutrient input and average model output. Both models had similar sensitivities to ration nutrient changes, and there was very little difference in model responses among rations, thereby allowing sensitivities to be pooled among rations. Animal outputs with greater sensitivities indicated more sensitive model outputs to changes in ration nutrient changes. For both models, predicted animal outputs were most sensitive to changes in CP, NDF, and SCP (lignin in CPM Dairy only; ether extract in Molly only) supporting the need for chemical analysis of these nutrients. Sensitivity of modelpredicted animal outputs to Met and Lys were ration and model dependent. Molly was sensitive to an AA, even if it was not limiting animal production, due to N excretion in milk and urine. Estimated AA allowable milk in CPM Dairy was the only output that was sensitive to a limiting AA. Both models were insensitive to changes in SC and ADF, indicating that model library values are adequate for ration formulation, evaluation, or both. Results suggest for both metabolic models a large number of ration input nutrients need not be chemically analyzed because they have no substantive impact on animal responses predicted by these models with the rations selected.