Improved prediction by enteric methane emission models in ruminant production systems by integrating climate classification

Abstract

Several mathematical models predict enteric methane (CH4) production (g/d) from ruminants. These models were developed from data originated from different regions and production systems and are not always applicable elsewhere. This study evaluated the performance of selected existing models for predicting enteric CH4 emissions of ruminants raised in diverse production systems in Argentina, Brazil, Chile and Uruguay. Climate was classified using the Köppen climate classification, while diets were classified by composition, such as high- or low-NDF, ether extract (EE), starch content and NDF digestibility. These classifications were applied to the diets fed to cattle (dairy, beef) and sheep. Models were evaluated and ranked by the lowest root mean square prediction error (RMSPE) and the ratio of RMSPE to SD of observed values (RSR). Models from the Intergovernmental Panel on Climate Change and those developed with Latin American data were used as reference. In temperate, no dry season, hot summer climates, all models performed poorly (RSR > 1) for dairy cattle. By diet composition, four models performed well (RSR < 1) for high-NDF diets. For beef cattle, the model by Yan et al. (2009) developed for forage-fed beef cattle performed best (RSR = 0.85) as all diets had high NDF. For sheep, the model by Congio et al. (2022a), which incorporates DM intake (DMI) and feeding level, performed best (RSR = 0.63); however, for highNDF diets, the model from Belanche et al. (2023) performed best. In temperate, no dry season, warm summer (Cfb) climates, models by Mills et al. (2003), which included DMI, performed best for dairy cattle (RSR = 0.78), including when assessed by diet composition. For low-EE diets, CH4 production (g/d) was best predicted using models with NDF intake (NDFI) and EE. For beef cattle, van Lingen et al. (2019), which included DMI, had the lowest RSR (0.91). By diet composition, models integrating fatty acids (FA), DMI, and NDFI performed best for low-NDF diets, while FA- and DMI-based models were most accurate for high-NDF diets. For sheep in Cfb climate, all models performed poorly. In tropical savanna, dry winter climates, all models performed poorly for beef cattle; similarly, poor performance was obtained when assessed by diet compositions. In conclusion, to improve CH4 emission prediction, model development should consider the potential effect of climate zones, together with ruminant category, feed intake, dietary composition (including potential of mitigation strategies), and feed digestibility parameters.