1 Department of Animal Science - Animal nutrition and physiology, Department of Animal Science, Science and Technology, Aarhus University2 Aberystwyth University3 Department of Animal Science - Animal nutrition and physiology, Department of Animal Science, Science and Technology, Aarhus University
This study explored the potential of partial least squares (PLS) and Fourier-transform infrared spectroscopy (FTIR) to predict rumen dry matter (DM) and neutral detergent fiber (NDF) degradation parameters of a wide range of feeds for ruminants, as an alternative to the in situ method. In total, 663 samples comprising 80 different feed types were analyzed. In situ DM and NDF degradabilities were determined as follows: effective degradability (ED), rumen soluble fraction (A), degradable but not soluble fraction (B), rate of degradation of the B fraction (C), and indigestible NDF (iNDF). Infrared spectra of dry samples were collected by attenuated total reflectance from 600 to 4000 cm−1. Feeds were randomly classified into 2 subsets of samples with representation of all feed types; one subset was used to develop regression models using partial least squares, and the second subset was used to conduct an external validation of the models. This study indicated that universal models containing all feed types and specific models containing concentrate feeds could provide only a relatively poor estimation of in situ DM degradation parameters because of compositional heterogeneity. More research, such as a particle size distribution analysis, is required to determine whether this lack of accuracy was due to limitations of the FTIR approach, or simply due to methodological error associated with the in situ method. This latter hypothesis may explain the low accuracy observed in the prediction of degradation rates if there was physical leakage of fine particles from the mesh bags used during in situ studies. In contrast, much better predictions were obtained when models were developed for forage feeds alone. Models for forages led to accurate predictions of DMA, DMB, NDFED, and NDF concentration (R2 = 0.91, 0.89, 0.85, and 0.79, standard error = 4.34, 5.97, 4.59, and 4.41% of DM, respectively), and could be used for screening of DMED, NDFC, and iNDF. These models relied on certain regions of the FTIR spectrum (900–1150 and 1500–1700 cm−1), which are mainly compatible with absorption of plant cell wall components, such as cellulose, pectin, lignin, cutin, and suberin, but also with nonstructural carbohydrates and certain active compounds. In conclusion, FTIR spectroscopy could be considered a low-cost alternative to in situ measurements in feed evaluation.
Journal of Dairy Science, 2014, Vol 97, Issue 4, p. 2361-2375