Predicting enteric methane emission of dairy cows with milk Fourier-transform infrared spectra and gas chromatography-based milk fatty acid profiles

The objective of the present study was to compare the prediction potential of milk Fourier-transform infrared spectroscopy (FTIR) for CH4 emissions of dairy cows with that of gas chromatography (GC)-based milk fatty acids (MFA). Data from 9 experiments with lactating Holstein-Friesian cows, with a total of 30 dietary treatments and 218 observations, were used. Methane emissions were measured for 3 consecutive days in climate respiration chambers and expressed as production (g/d), yield (g/kg of dry matter intake; DMI), and intensity (g/kg of fat- and protein-corrected milk; FPCM). Dry matter intake was 16.3 +/- 2.18 kg/d (mean +/- standard deviation), FPCM yield was 25.9 +/- 5.06 kg/d, CH4 production was 366 +/- 53.9 g/d, CH4 yield was 22.5 +/- 2.10 g/kg of DMI, and CH4 intensity was 14.4 +/- 2.58 g/kg of FPCM. Milk was sampled during the same days and analyzed by GC arid by FTIR. Multivariate GC-determined MFA-based and FTIR-based CH4 prediction models were developed, and subsequently, the final CH4 prediction models were evaluated with root mean squared error of prediction and concordance correlation coefficient analysis. Further, we performed a random 10-fold cross validation to calculate the performance parameters of the models (e.g., the coefficient of determination of cross validation). The final GC-determined MFA-based CH4 prediction models estimate CH4 production, yield, and intensity with a root mean squared error of prediction of 35.7 g/d, 1.6 g/kg of DMI, and 1.6 g/kg of FPCM and with a concordance correlation coefficient of 0.72, 0.59, and 0.77, respectively. The final FTIR-based CH4 prediction models estimate CH4 production, yield, and intensity with a root mean squared error of prediction of 43.2 g/d, 1.9 g/kg of DMI, and 1.7 g/kg of FPCM and with a concordance correlation coefficient of 0.52, 0.40, and 0.72, respectively. The GC-determined MFA-based prediction models described a greater part of the observed variation in CH4 emission than did the FTIR-based models. The cross validation results indicate that all CH4 prediction models (both GC-determined MFA based arid FTIR-based models) are robust; the difference between the coefficient of determination and the coefficient of determination of cross validation ranged from 0.01 to 0.07. The results indicate that GC-determined MFA have a greater potential than FTIR spectra to estimate CH4 production, yield, and intensity. Both techniques hold potential but may not yet be ready to predict CH4 emission of dairy cows in practice. Additional CH4 measurements are needed to improve the accuracy and robustness of GC-determined MFA and FTIR spectra for CH4 prediction.