Is it possible to predict the methane emission intensity of Swedish dairy cows from milk spectra?

Abstract

Emissions of greenhouse gases (GHG), especially methane (CH4), from dairy production have received much research attention in the past 15 years, with the main focus being to identify factors affecting CH4 production and measures to reduce CH4 emissions from dairy cows. However, measurement of CH4 production by dairy cows in commercial herds is time-consuming and requires expensive equipment, so there is a need to find alternative ways to estimate individual and herd CH4 emissions. Regular milk analyses are performed for many cows, so data from mid-infrared spectroscopy (MIRS) on individual milk samples could perhaps be utilised to predict CH4 emissions intensity (MI, CH4 g/kg milk production). This study investigated the potential and limitations of predicting individual MI by integrating data from CH4 measurements made by an infrared sniffer (IRS) and milk MIRS data taken from fortnightly morning milkings during the full lactation records of 37 multiparous cows. Partial least square regression was used to create prediction models for six-week lactation sub-periods and for full lactations, which were validated using leave-one-cow-out cross-validation. Coefficient of determination in predicting MI was low, indicating that the method is not suitable for predicting variations in individual MI, although it should further be evaluated at herd level.

Keywords

• Infrared sniffers
• Partial least square regression
• Prediction
• MIRS