The AusBeef model for beef production: I. Description and evaluation

Abstract

As demand for animal products, such as meat and milk, increases, and concern over environmental impact grows, mechanistic models can be useful tools to better represent and understand ruminant systems and evaluate mitigation options to reduce greenhouse gas emissions without compromising productivity. The objectives of the present study were to describe the representation of processes for growth and enteric methane (CH₄) production in AusBeef, a whole-animal, dynamic, mechanistic model for beef production; evaluate AusBeef for its ability to predict daily methane production (DMP, g/day), gross energy intake (GEI, MJ/day) and methane yield (MJ CH₄/MJ GEI) using an independent data set; and to compare AusBeef estimates to those from the empirical equations featured in the current National Academies of Sciences, Engineering and Medicine (NASEM, 2016) beef cattle requirements for growth and the Ruminant Nutrition System (RNS), a dynamic, mechanistic model of Tedeschi & Fox, 2016. AusBeef incorporates a unique fermentation stoichiometry that represents four microbial groups: protozoa, amylolytic bacteria, cellulolytic bacteria and lactate-utilizing bacteria. AusBeef also accounts for the effects of ruminal pH on microbial degradation of feed particles. Methane emissions are calculated from net ruminal hydrogen balance, which is defined as the difference between inputs from fermentation and outputs due to microbial use and biohydrogenation. AusBeef performed similarly to the NASEM empirical model in terms of prediction accuracy and error decomposition, and with less root mean square predicted error (RMSPE) than the RNS mechanistic model when expressed as a percentage of the observed mean (RMSPE, %), and the majority of error was non-systematic. For DMP, RMSPE for AusBeef, NASEM and RNS were 24·0, 19·8 and 50·0 g/day for the full data set (<i>n</i> = 35); 25·6, 18·2 and 56·2 g/day for forage diets (<i>n</i> = 19); and 21·8, 21·5 and 41·5 g/day for mixed diets (<i>n</i> = 16), respectively. Concordance correlation coefficients (CCC) were highest for GEI, with all models having CCC > 0·66, and higher CCC for forage diets than mixed, while CCC were lowest for MY, particularly forage diets. Systematic error increased for all models on forage diets, largely due to an increase in error due to mean bias, and while all models performed well for mixed diets, further refinements are required to improve the prediction of CH₄ on forage diets.