Evaluation of AusBeef Methane Predictions from Beef Cattle

Abstract

The AusBEEF model (version 1.2) has been evaluated against data from 5 experiments (12 experimental means) with beef cattle which had methane emissions measured in respiration chambers. The dataset included a range of animal ages, gender and diet types (perennial ryegrass ranging from vegetative to mature, mixes of ryegrass with maize silage or palm kernel, maize silage alone, and forage rape). The predictions of the model for absolute methane emissions (g/d) and relative methane emissions (yield per unit of DM intake: yCH4 and as a percentage of gross energy intake: CH4%GEI) were evaluated against the observed values. Irrespective of the unit used for methane emissions, AusBEEF mean predictions were slightly (~5%) greater than the observed data. The agreement between observed and predicted values was good for absolute methane emissions (concordance correlation coefficient 0.86) and moderate for yield measurements (CCC 0.48 and 0.58 for yCH4 and CH4%GEI, respectively). For yield measurements, a systematic slope bias accounted for ~40% of the mean square prediction error (MSPE). The ratio of the root MSPE to standard deviation of the observed values (RSR), was used to assess the model predictions in context to the inherent variability of the observed data. Based on the RSR, AusBeef predicted absolute methane emissions very well (RSR=0.5), but prediction of relative methane emissions could be improved (RSR 1.0 for yCH4 and CH4%GEI, respectively). AusBEEF predictions were correctly ranked for forage rape and mixes of pasture and supplements. However, the predictions were not as for diets of 100% pasture. AusBeef predicted live weight losses for most experiments in the database, in contrast with the observed data. These results suggest that improvements in the representation of digestive processes may be required in AusBEEF if accurate predictions of both methane and animal performance are to be obtained, particularly for forage-only diets.