Evaluation of extant models for predicting methane emissions in sheep fed forage from Campos Rangeland

Abstract

Predictive models have been developed to estimate methane (CH₄) emissions from ruminant feeding strategies. However, their accuracy can decline when applied across diverse production systems. This issue is relevant in countries like Uruguay, where agroclimatic conditions, feeding strategies, and livestock practices differ from those represented in global model datasets. To address this, we evaluated the performance of existing models in predicting enteric CH₄ production (g/day) using a regional dataset from sheep fed forage from Uruguayan rangeland. Model selection followed PRISMA guidelines, focusing on applicability across agroclimatic regions and sheep-specific systems. The IPCC (2019) model was used as a reference, while Congio et al. (2022) was selected because it was developed using Latin America data. Other models included Ramin (2013), Patra (2016), and Belanche et al. (2023) were also tested. Models were applied in a dataset from an experiment conducted at INIA Glencoe Experimental Station. Sixteen Corriedale lambs were individually housed in metabolic cages in a completely randomized design and assigned to increasing amounts of forage obtained from native rangeland (Tafernaberry et al., 2024) over two 20-day periods (n=32). CH₄ emissions were measured using the PAC system at the end of each period. Descriptive statistics were performed on independent variables: dry matter intake (DMI), and body weight and the dependent variable CH₄ emissions. Observed CH₄ emissions averaged 18.8g/day (SD=2.1), with DMI averaging 0.89 kg/sheep/day (SD=0.2). Model evaluation employed a mixed linear model and was assessed using Root Mean Squared Prediction Error (RMSPE) and the Concordance Correlation Coefficient. Among the models tested, Belanche (2023), which incorporates DMI and body weight, demonstrated the best performance (RMSPE = 7.6%; CCC=0.73), outperforming the IPCC (2019) model (RMSPE=16.4%; CCC=0.47), which overpredicted CH₄ emissions by 2.5g/sheep/day. These findings highlight the value of region-specific models in improving the accuracy of enteric CH₄ emission estimation in different rangeland regions.