Evaluation of the performance of existing mathematical models predicting enteric methane emissions from ruminants:Animal categories and dietary mitigation strategies

The objective of this study was to evaluate the performance of existing models predicting enteric methane (CH) emissions, using a large database (3183 individual data from 103 in vivo studies on dairy and beef cattle, sheep and goats fed diets from different countries). The impacts of dietary strategies to reduce CH emissions, and of diet quality (described by organic matter digestibility (dOM) and neutral-detergent fiber digestibility (dNDF)) on model performance were assessed by animal category. The models were first assessed based on the root mean square prediction error (RMSPE) to standard deviation of observed values ratio (RSR) to account for differences in data between models and then on the RMSPE. For dairy cattle, the CH (g/d) predicting model based on feeding level (dry matter intake (DMI)/body weight (BW)), energy digestibility (dGE) and ether extract (EE) had the smallest RSR (0.66) for all diets, as well as for the high-EE diets (RSR = 0.73). For mitigation strategies based on lowering NDF or improving dOM, the same model (RSR = 0.48 to 0.60) and the model using DMI and neutral- and acid-detergent fiber intakes (RSR = 0.53) had the smallest RSR, respectively. For diets with high starch (STA), the model based on nitrogen, ADF and STA intake presented the smallest RSR (0.84). For beef cattle, all evaluated models performed moderately compared with the models of dairy cattle. The smallest RSR (0.83) was obtained using variables of energy intake, BW, forage content and dietary fat, and also for the high-EE and the low-NDF diets (RSR = 0.84 to 0.86). The IPCC Tier 2 models performed better when dietary STA, dOM or dNDF were high. For sheep and goats, the smallest RSR was observed from a model for sheep based on dGE intake (RSR = 0.61). Both IPCC models had low predictive ability when dietary EE, NDF, dOM and dNDF varied (RSR = 0.57 to 1.31 in dairy, and 0.65 to 1.24 in beef cattle). The performance of models depends mostly on explanatory variables and not on the type of data (individual vs. treatment means) used in their development or evaluation. Some empirical models give satisfactory prediction error compared with the error associated with measurement methods. For better prediction, models should include feed intake, digestibility and additional information on dietary concentrations of EE and structural and nonstructural carbohydrates to account for different dietary mitigating strategies.
This study is part of the Joint Programming Initiative on Agriculture, Food Security and Climate Change (FACCE-JPI)’s “GLOBAL NETWORK” project and the “Feeding and Nutrition Network” (http://animalscience.psu.edu/fnn) of the Livestock Research Group within the Global Research Alliance for Agricultural Greenhouse Gases (www.globalresearchalliance.org). Authors gratefully acknowledge funding for this project from: USDA National Institute of Food and Agriculture (Grant no. 2014-67003-21979) University of California, Davis Sesnon Endowed Chair Program, USDA, and Austin Eugene Lyons Fellowship (University of California, Davis); Funding from USDA National Institute of Food and Agriculture Federal Appropriations under Project PEN 04539 and Accession number 1000803, DSM Nutritional Products (Basel, Switzerland), Pennsylvania Soybean Board (Harrisburg, PA, USA), Northeast Sustainable Agriculture Research and Education (Burlington, VT, USA), and PMI Nutritional Additives (Shoreview, MN, USA); the Ministry of Economic Affairs (the Netherlands; project BO-20-007-006; Global Research Alliance on Agricultural Greenhouse Gases), the Product Board Animal Feed (Zoetermeer, the Netherlands) and the Dutch Dairy Board (Zoetermeer, the Netherlands); The Cofund for Monitoring & Mitigation of Greenhouse Gases from Agri- and Silvi-culture (FACCE ERA-GAS)’s project Capturing Effects of Diet on Emissions from Ruminant Systems and the Dutch Ministry of Agriculture, Nature and Food Quality (AF-EU-18010 & BO-4400159-01); USDA National Institute of Food and Agriculture (Hatch Multistate NC-1042 Project Number NH00616-R; Project Accession Number 1001855) and the New Hampshire Agricultural Experiment Station (Durham, NH); French National Research Agency through the FACCE-JPI program (ANR-13-JFAC-0003-01); the Department of Agriculture, Food and the Marine, Ireland Agricultural GHG Research Initiative for Ireland (AGRI-I) project; Academy of Finland (No. 281337), Helsinki, Finland; Swiss Federal Office of Agriculture, Berne, Switzerland; the Department for Environment, Food and Rural Affairs (Defra; UK); Defra, the Scottish Government, DARD, and the Welsh Government as part of the UK’s Agricultural GHG Research Platform projects (www.ghgplatform.org.uk); INIA (Spain, project MIT01-GLOBALNET-EEZ); German Federal Ministry of Food and Agriculture (BMBL) through the Federal Office for Agriculture and Food (BLE); Swedish Infrastructure for Ecosystem Science (SITES) at Röbäcksdalen Research Station; Comisión Nacional de Investigación Científica y Tecnológica, Fondo Nacional de Desarrollo Científico y Tecnológico (Grant Nos. 11110410 and 1151355) and Fondo Regional de Tecnología Agropecuaria (FTG/RF-1028-RG); European Commission through SMEthane (FP7-SME-262270).