1st Page



This article in

  1. Vol. 87 No. 4, p. 1334-1345
    Received: Nov 13, 2007
    Accepted: Dec 03, 2008
    Published: December 5, 2014

    2 Corresponding author(s):


Modeling methane production from beef cattle using linear and nonlinear approaches1

  1. J. L. Ellis*2,
  2. E. Kebreab,
  3. N. E. Odongo*‡,
  4. K. Beauchemin§,
  5. S. McGinn§,
  6. J. D. Nkrumah#,
  7. S. S. Moore#,
  8. R. Christopherson#,
  9. G. K. Murdoch#||,
  10. B. W. McBride*,
  11. E. K. Okine# and
  12. J. France*
  1. Centre for Nutrition Modeling, Department of Animal and Poultry Science, University of Guelph, Guelph, Ontario, N1G 2W1, Canada;
    National Centre for Livestock and Environment, Department of Animal Science, University of Manitoba, Winnipeg, Manitoba, R3T 2N2, Canada;
    Animal Production and Health Section, Department of Nuclear Sciences and Applications, International Atomic Energy Agency, PO Box 100, Wagramer Strasse 5, A-1400 Vienna, Austria;
    Agriculture and Agri-Food Canada, Lethbridge Research Centre, Lethbridge, AB, T1J 4B1, Canada;
    Department of Agricultural, Food and Nutritional Science, University of Alberta, Edmonton, Alberta, T6G 2P5, Canada; and
    Animal and Veterinary Science, University of Idaho, Moscow 83844


Canada is committed to reducing its greenhouse gas emissions to 6% below 1990 amounts between 2008 and 2012, and methane is one of several greenhouse gases being targeted for reduction. Methane production from ruminants is one area in which the agriculture sector can contribute to reducing our global impact. Through mathematical modeling, we can further our understanding of factors that control methane production, improve national or global greenhouse gas inventories, and investigate mitigation strategies to reduce overall emissions. The purpose of this study was to compile an extensive database of methane production values measured on beef cattle, and to generate linear and nonlinear equations to predict methane production from variables that describe the diet. Extant methane prediction equations were also evaluated. The linear equation developed with the smallest root mean square prediction error (RMSPE, % observed mean) and residual variance (RV) was Eq. I: CH4, MJ/d = 2.72 (±0.543) + [0.0937 (±0.0117) × ME intake, MJ/d] + [4.31 (±0.215) × Cellulose, kg/d] − [6.49 (±0.800) × Hemicellulose, kg/d] − [7.44 (±0.521) × Fat, kg/d] [RMSPE = 26.9%, with 94% of mean square prediction error (MSPE) being random error; RV = 1.13]. Equations based on ratios of one diet variable to another were also generated, and Eq. P, CH4, MJ/d = 2.50 (±0.649) − [0.367 (±0.0191) × (Starch:ADF)] + [0.766 (±0.116) × DMI, kg/d], resulted in the smallest RMSPE values among these equations (RMSPE = 28.6%, with 93.6% of MSPE from random error; RV = 1.35). Among the nonlinear equations developed, Eq. W, CH4, MJ/d = 10.8 (±1.45) × (1 − e[−0.141 (±0.0381) × DMI, kg/d]), performed well (RMSPE = 29.0%, with 93.6% of MSPE from random error; RV = 3.06), as did Eq. W3, CH4, MJ/d = 10.8 (±1.45) × [1 − e{− [−0.034 × (NFC/NDF) + 0.228] × DMI, kg/d}] (RMSPE = 28.0%, with 95% of MSPE from random error). Extant equations from a previous publication by the authors performed comparably with, if not better than in some cases, the newly developed equations. Equation selection by users should be based on RV and RMSPE analysis, input variables available to the user, and the diet fed, because the equation selected must account for divergence from a “normal” diet (e.g., high-concentrate diets, high-fat diets).

Copyright © 2009. Copyright 2009 Journal of Animal Science