Potential of faecal NIRS for predicting methane emissions in dairy cows supplemented with 3-nitrooxypropanol

Current methods used to measure methane (CH₄) emissions from individual ruminants are difficult to apply on farms. Alternative techniques, particularly spectroscopic methods, offer a strong potential to predict individual CH₄ emissions across diverse ruminants under farm conditions. The aims of this work were to i) evaluate the potential of faecal NIRS to detect the effects of 3-nitrooxypropanol (3-NOP) as an antimethanogenic additive to dairy cows' diets in comparison with milk mid-infrared (MIR) spectroscopy and ii) test preliminary models of CH₄ emissions from faecal NIRS of cows supplemented or not with 3-NOP. This work was based on two previous in vivo experiments conducted under controlled conditions in dairy cows (n = 43 in total) fed different diets, which evaluated the effect of 3-NOP (60 mg/kg DM) on CH₄ emissions. Individual data on CH₄ were acquired using GreenFeed (Experiment 1) or respiration chambers (Experiment 2). Individual milk and faecal samples for MIR spectroscopy and NIRS analyses, respectively, were collected during the same week of CH₄ measurement. Observed CH₄ emissions (g/day) were significantly (P < 0.01) lower by 32 and 23% in cows fed 3-NOP-supplemented diets in Experiments 1 and 2, respectively. Milk MIR spectra did not detect cows fed with or without 3-NOP, whereas faecal NIRS did. When an existing local NIRS model for predicting CH₄ emissions was applied to faecal spectra from Experiment 1, the predicted values for 3-NOP spectra did not align with the observed measurements. The root mean square error of prediction was 174.7 g/day for the 3-NOP treatment and 75.7 g/day for the control treatment. These values were mainly explained by the bias (160.9 vs -3.9 g/day, for the 3-NOP and control treatments, respectively). To improve the CH₄ emission prediction of 3-NOP treatment, a specific exploratory prediction model was developed posthoc for 3-NOP faecal samples in each experiment. The R² values and the root mean square errors of cross-validation were 0.7 and 21.7 g/day for Experiments 1 and 2, respectively. This preliminary work suggests the potential of faecal NIRS to predict CH₄ emissions from dairy cows fed diets supplemented with 3-NOP when using specific models. Further work is needed to confirm this proof of concept, calibrate specific models with additional 3-NOP data, and extend this approach to other antimethanogenic additives in future on-farm applications.