The use of mid-infrared spectrometry to estimate the ration composition of lactating dairy cows

Summary in the original language of the document

The composition of cow milk is strongly affected by the feeding regimen. Because milk components are routinely determined using mid-infrared (MIR) spectrometry, MIR spectra could also be used to estimate an animal’s ration composition. The objective of this study was to determine whether and how well amounts of dry matter intake and the proportions of concentrates, hay, grass silage, maize silage, and pasture in the total ration can be estimated using MIR spectra at an individual animal level. A total of 10,200 milk samples and sets of feed intake data were collected from 90 dairy cows at 2 experimental farms of the Agricultural Research and Education Centre in Raumberg-Gumpenstein, Austria. For each run of analysis, the data set was split into a calibration and a validation data set in a 40:60 ratio. Estimated ration compositions were calculated using a partial least squares regression and then compared with the respective observed ration compositions. In separate analyses, the factors milk yield and concentrate intake were included as additional predictors. To evaluate accuracy, the coefficient of determination (R2) and ratio to performance deviation were used. The highest R2 values (for kg of dry matter intake/ for % of ration) for the individual feedstuffs were as follows: pasture, 0.63/0.66; grass silage, 0.32/0.43; concentrate intake, 0.39/0.34; maize silage, 0.32/0.33; and hay, 0.15/0.16. Estimation of groups of feedstuffs (forages, energy-dense feedstuffs) mostly resulted in R2 values >0.50. Including the parameters milk yield or concentrate intake improved R2 values by up to 0.21, with an average improvement of 0.04. The results of this study indicate that not all ration components may be estimated equally accurately. Even if some estimates are good on average, there may be strong deviations between estimated and observed values in individual data sets, and therefore individual estimates should not be overemphasized. Further research including pooled samples (e.g., bulk milk, farm samples) or variations in ration composition is called for.