Accurate estimates of methane (CH4) production by cattle in different contexts are essential to developing mitigation strategies in different regions. We aimed to: (i) compile a database of CH4 emissions from Brazilian cattle studies, (ii) evaluate prediction precision and accuracy of extant proposed equations for cattle and (iii) develop specialized equations for predicting CH4 emissions from cattle in tropical conditions. Data of nutrient intake, diet composition and CH4 emissions were compiled from in vivo studies using open-circuit respiratory chambers, SF6 technique or the GreenFeed® system. A final dataset containing intake, diet composition, digestibility and CH4 emissions (677 individual animal observations, 40 treatment means) obtained from 38 studies conducted in Brazil was used. The dataset was divided into three groups: all animals (GEN), lactating dairy cows (LAC) and growing cattle and non-lactating dairy cows (GCNL). A total of 54 prediction equations available in the literature were evaluated. A total of 96 multiple linear models were developed for predicting CH4 production (MJ/day). The predictor variables were DM intake (DMI), gross energy (GE) intake, BW, DMI as proportion of BW, NDF concentration, ether extract (EE) concentration, dietary proportion of concentrate and GE digestibility. Model selection criteria were significance (P < 0.05) and variance inflation factor lower than three for all predictors. Each model performance was evaluated by leave-one-out cross-validation. The Intergovernmental Panel on Climate Change (2006) Tier 2 method performed better for GEN and GCNL than LAC and overpredicted CH4 production for all datasets. Increasing complexity of the newly developed models resulted in greater performance. The GCNL had a greater number of equations with expanded possibilities to correct for diet characteristics such as EE and NDF concentrations and dietary proportion of concentrate. For the LAC dataset, equations based on intake and animal characteristics were developed. The equations developed in the present study can be useful for accurate and precise estimation of CH4 emissions from cattle in tropical conditions. These equations could improve accuracy of greenhouse gas inventories for tropical countries. The results provide a better understanding of the dietary and animal characteristics that influence the production of enteric CH4 in tropical production systems.

在不同情况下对牛甲烷（CH 4）产量的准确估算对于制定不同地区的缓解策略至关重要。我们旨在：（i）汇编巴西牛研究中的CH 4排放数据库，（ii）评估现有提议的牛方程的预测精度和准确性，（iii）开发用于预测热带条件下牛CH 4排放的专门方程。营养摄入，饮食成分和数据CH 4排放量从已编译的体内研究利用开路呼吸室，SF 6技术或绿色饲料®系统。使用从巴西进行的38项研究获得的最终数据集，包括摄入量，饮食组成，消化率和CH 4排放量（677个动物个体观察值，40种治疗手段）。数据集分为三组：所有动物（GEN），泌乳奶牛（LAC）和生长中的奶牛和非泌乳奶牛（GCNL）。文献中总共评估了54个预测方程。总共开发了96个用于预测CH 4产量（MJ /天）的线性模型。预测变量为DM摄入量（DMI），总能量（GE）摄入量，BW，DMI占BW的比例，NDF浓度，乙醚提取物（EE浓度，日粮中精矿的比例和GE的消化率。模型选择标准具有显着性（P <0.05），并且所有预测变量的方差膨胀因子均低于3。通过留一法交叉验证评估每种模型的性能。政府间气候变化专门委员会（2006）的方法2对GEN和GCNL的效果优于LAC，并且CH 4的预测值过高所有数据集的生产。新开发模型的复杂性不断提高，从而带来了更高的性能。GCNL具有更多的方程式，并且具有扩展的可能性来校正饮食特征，例如EE和NDF浓度以及浓缩饲料的饮食比例。对于LAC数据集，开发了基于摄入量和动物特征的方程式。在本研究中开发的方程式可用于准确和精确地估计热带条件下牛的CH 4排放。这些等式可以提高热带国家温室气体清单的准确性。结果提供了对饮食和动物特征的更好理解，这些饮食和动物特征会影响热带生产系统中肠道CH 4的产生。