Prediction and evaluation of enteric methane emissions from lactating dairy cows using different levels of covariate information

Abstract

The dairy sector contributes to global warming through enteric methane (CH4) emissions. Methane is also a loss of energy to the ruminant. Several studies have developed CH4 prediction models to assess mitigation strategies to reduce emissions. However, the majority of these models have low predictive ability or require numerous inputs that are often not readily available in commercial dairy operations. In this context, the objective of the present paper was to develop CH4 prediction models by using varying levels of information available at the farm level. The seven complexity levels used the following information: (1) dietary nutrient composition, (2) milk yield and composition, (3) Levels 1 and 2, (4) Level 3 plus dry matter intake (DMI), (5) Level 4 plus bodyweight, (6) Level 2 plus DMI, and (7) DMI only. Models were fitted to 489 individual enteric-CH4 measurements from 30 indirect calorimetry studies and evaluated with an independent database comprising 215 treatment means from 62 studies collected from the literature. Within each complexity level, all possible mixed-effect models were fitted and those with the lowest values of Akaike or Bayesian information criteria were selected using lme4 package in R. Models were evaluated using mean square prediction error (MSPE) based statistic, root MSPE (RMSPE) to observation standard deviation ratio, concordance correlation coefficient and Nash–Sutcliffe efficiency methods. All fitted models performed well with an acceptable error estimates (RMSPE as a percentage of observed mean (RMSPE%) = 16–24%), with more than two-thirds of total error originating from random bias. Overall, models with DMI were more accurate (RMSPE% = 16–20%) than those without (RMSPE% = 20–24%). Although the best prediction model (RMSPE% = 16%) was developed using Level 5 information, a model using Level 2 information is recommended for on-farm methane estimates if DMI is not measured. The proposed models offer easy and practical tools to dairy producers for predicting CH4 emissions and evaluating CH4 mitigation strategies.