A simulation model “CTR Dairy” to predict the supply of nutrients in dairy cows managed under discontinuous feeding patterns

Abstract

A simulation rumen model has been developed to function under non-steady state conditions in order to allow prediction of nutrient availability in dairy cows managed under discontinuous feeding systems. The model simulates availability of glycogenic, aminogenic and lipogenic nutrients to lactating dairy cows fed discontinuously. The model structure considers input of up to three different feeds fed independently at any time during the day. Feeds are described by their nitrogen (N), carbohydrate and fatty acid fractions. The N containing feed fractions include ruminally undegraded crude protein (CP), ruminally insoluble but potentially degradable CP, ruminally soluble CP and ammonia N. The feed carbohydrate fractions include ruminally undegradable neutral detergent fibre (NDF), ruminally degradable NDF, ruminally insoluble starch, ruminally soluble starch and sugars. The fatty acids in the feeds are divided between long chain fatty acids and volatile fatty acids (VFA). Additionally four pools were defined representing absorption of amino acids, glucose, long chain fatty acids and volatile fatty acids. The rumen microbial population is represented as a single pool. Besides a flexible structure, new features to the extant model include adoption of the concept of chewing efficiency (or chewing effectiveness) during eating, variable fractional ruminal absorption rates of VFA and variable fractional ruminal degradation rates of NDF as a function of rumen liquid pH, as well as a variable rumen volume which directly affects rumen concentrations of metabolites. The model continuously ( i.e., by minute) predicts release of soluble components from the feeds in the rumen, concentration and absorption of fermentation end products in the rumen, rumen pools of nutrients and microbial biomass dynamics, as well as passage of microbial biomass and non-fermented nutrients from the rumen, in response to various feeding strategies. Model evaluation covered a wide range of feeding strategies that included pasture and housed feeding systems. Overall, the mean square prediction error (MSPE) as a percentage of the observed mean was relatively low (<10%) with a high amount of the total variation explained by random variation (>65%). Deviation from unity varied between 23% (rumen dry matter content) and 25% (NDF), indicating some consistent over and/or under prediction. A more detailed evaluation was done based on studies available that reported diurnal behaviour of key model outputs such as rumen pools, rumen pH, and rumen VFA. The predictions broadly simulated the observed values quantitatively, relative to general diurnal patterns, and relative to differences between treatments in the predicted diurnal patterns. Results show that the model provides a tool to assess potential outcomes of changing feeding strategies which may be particularly valuable in assessing selection of feeds, amounts and times of the day to offer the feeds. The continuous nature of the simulated output also allows determination of the time(s) of the day that ruminal (and/or post-ruminal) delivery of nutrients may limit ruminal output of nutrients (and/or availability of nutrients) to support milk nutrient synthesis.