Estimation of methane eructed by dairy and beef cattle using faecal near-infrared spectra

Session 28. SmartCow: Methods and technologies for research and smart nutrition management in dairy and beef cattlePrésentation Orale

International audience

إنجليزي. The relevance to consider faecal near-infrared spectra (NIRS) to predict digestible organic matter (DOM) has beenestablished while methane (CH4) eructation, is also related, among others, to the intake of DOM. Consequently, itseems also relevant to investigate the estimation of eructed CH4 from faecal NIRS. A reference dataset collected indifferent countries (Belgium, France and Switzerland), including faecal NIRS and corresponding CH4 values, hasbeen constituted in the framework of the Smartcow project (H2020 N°730924). Automated Head-Chamber System(AHCS, GreenFeed) was used to collect CH4 production references data. Only values including at least 20 visits ofthe AHCS during the period were considered as representative. A period of 3 weeks for CH4 measurement has beenfixed to maximize the dataset. In parallel spot samples of faeces were collected at the end of the period. Data fromlactating cows (LC) [n=91, 361±86 g CH4/day (mean ± SD), Holstein LC, diet based on grass silage, grass and/orhay] were considered separately from beef cattle (BC) data [n=346, 221±43 g CH4/day, Charolais (heifer), BelgianBlue (reformed cow), Dual purpose Belgian Blue (suckling cow), diet based on grass, grass silage and/or hay]. Fromthe data available, the best models presented for LC and BC respectively, a R2 of calibration about 0.43 and 0.62,and a standard error of calibration about 65 and 26 g CH4/day. The better results for BC than for LC were expectedbecause this calibration set include data from more animals and conditions, consequently more variability at animallevel but also regarding CH4 and spectral information, which is needed to develop such models. These results suggestthe feasibility of having a proxy based on faecal NIRS to estimate CH4 eructated which is particularly interestingfor non-lactating animals (milk-based proxies are inapplicable). Additional data have to be collected with the samesampling protocol but presenting other diets, breeds, physiological status etc. to confirm this trend and to improvethe model robustness.