The objective of this study was to develop linear and nonlinear statistical models for predicting enteric methane emissions from beef and dairy cattle (EME, MJ/day). Ration nutrient composition (g/kg), nutrient (kg/day), energy (MJ/day), and energy and organic matter (OM) digestibility (g/kg) were used as predictors of CH4 production. Three databases of beef cattle, dairy cattle, and their combinations were developed using 34 published experiments to model EME predictions. Linear and nonlinear regression models were developed using a mixed-model approach to predict CH4 production (MJ/day) of individual animals based on feed composition. For the beef cattle database, Equation methane (MJ/d) = 1.6063 (±0.757) + 0.4256 (±0.0745) × DMI + 1.2213 (±0.1715) × NDFI + -0.475 (±0.446) × ADFI had the smallest RMSPE (21.99%), with 83.51% of this coming from random error and a regression bias was 16.49%. For the dairy cattle database, the RMSPE was minimized (15.99%) for methane (MJ/d) = 0.3989 (±1.1073) + 0.8685 (±0.1585) × DMI + 0.6675 (±0.4264) × NDFI, of which 85.11% was from random error and the regression deviation was 14.89%. When the beef and dairy cattle databases were combined, the RMSPE was minimized (24.4%) for methane(MJ/d) = -0.3496 (±0.723) + 0.5941 (±0.0851) × DMI + 1.388 (±0.2203) × NDFI + -0.027 (±0.4223) × ADFI, of which 85.62% was from the random error and the regression bias was 14.38%. Among the nonlinear equations for the three databases, the DMI-based exponential model outperformed the other nonlinear models, but the predictability and goodness of fit of the equations did not improve compared to the linear model. The existing equations overestimate CH4 production with low accuracy and precision. Therefore, the equations developed in this study improve the preparation of methane inventories and thus improve the estimation of methane production in beef and dairy cattle.
Keywords: beef cattle; dairy cattle; energy; methane emission; statistical model.