Predicting enteric methane emission in sheep using linear and non-linear statistical models from dietary variables

Abstract

The objective of the present study was to develop linear and non-linear statistical models for prediction of enteric methane emission (EME) in sheep. A database from 80 publications, which included a total of 449 mean observations of EME measured on more than 1500 sheep, was constructed to develop prediction and evaluation of models of EME. Dietary nutrient composition (g/kg), nutrient or energy intake (kg/day or MJ/day) and digestibility (g/kg) of organic matter were used as predictors of EME (MJ/day). The dietary concentrations of neutral detergent fibre and crude protein, and feed intake, were 435 g/kg, 152 g/kg and 0.92 kg/day, respectively. The EME by sheep expressed as MJ/day and % of gross energy intake was 1.02 and 6.54, respectively. The simple linear equation that predicted EME with high precision and accuracy was EME = 0.208(±0.040) + 0.049(±0.0039) × gross energy intake (MJ/day), adjusted R2 = 0.86 with root mean-square prediction error of 22.7%, of which 93% was from random error and regression bias of 3.20%. Additions of dietary concentration of fibre and feeding level, and organic matter digestibility to the simple linear model improved the models. Among the non-linear equations developed, monomolecular model, i.e. EME = 5.699 (±1.94) – [5.699 (±1.94) – 0.133 (±0.047)] × exp[–0.021(±0.0071) × metabolisable energy intake (MJ/day)]; adjusted R2 = 0.90 and mean-square prediction error = 20.1%, with 96.3% random error, performed better than simple linear and other non-linear models. The equations developed in the present study will be useful for national methane inventory preparation, and for a better understanding of dietary factors influencing EME in sheep.