View graph of relations

By means of a scanning monochromator and partial least-squares analysis, near-IR reflectance spectroscopy (NIRS) calibrations were developed to predict the chemical composition and the energy value of two different sets of compound feeds for cattle. The first set contained 179 compounds of heterogeneous nature, for which the energy value was calculated from in vivo digestibility. The second set consisted of 163 commercial dairy compounds for which the energy value was predicted from enzymatic digestibility. Spectral and reference data were highly correlated for all parameters, except for ash. For the first set, the root mean square error of prediction (RMSEP) amounted to 0.5% (percentage units) for moisture, 1.4% on dry matter basis (DM) for crude protein, 1.4% on DM for crude fibre, 0.7% on DM for crude fat, 2.1% for in vivo digestibility, 0.37 MJ kg-1 DM for metabolizable energy (ME) and 0.27 MJ kg-1 DM for net energy lactation (NEL). Prediction of ME and NEL with regression equations based on enzymatic digestibility was somewhat more accurate, with an RMSEP of 0.30 MJ kg-1 DM and 0.22 MJ kg-1 DM, respectively. For the more homogeneous second set, the RMSEP was lower for all parameters: 0.3% for moisture, 1.1% on DM for crude protein, 0.9% on DM for crude fibre, 0.3% on DM for crude fat, 1.3% for enzymatic digestibility, 0.30 MJ kg-1 DM for ME and 0.18 MJ kg-1 DM for NEL. It was proved that calibrations derived from energy values predicted from enzymatic digestibility were hardly less accurate than those derived from in vivo data. Calibrations based on the currently used NIRS instrument with 19 filters gave similar results to those based on the full spectrum for moisture, crude protein and crude fat, were less accurate for crude fibre, and clearly failed for energy evaluation. It was concluded that NIRS can be used as a screening method for the control of the chemical composition and the energy value of compound feeds for cattle.
Original languageEnglish
JournalAnimal Feed Science and Technology
Issue number3-4
Pages (from-to)243-253
Number of pages11
Publication statusPublished - 1-Feb-1995

ID: 206158