Prediction of energy value and protein content of forages by near infrared reflectance spectroscopy
1988
Lindgren, E. (Sveriges Lantbruksuniv., Uppsala (Sweden). Inst. foer Husdjurens Utfodring och Vaard)
Near Infrared Reflectance Spectroscopy (NIRS) was used to predict the quality of grass, legume-grass mixtures and legumes. Altogether 440 samples of forage with known contents of metabolizable energy and crude protein were used to develop and evaluate calibrations for these parameters on a scanning instrument (Technicon InfraAlyzer 500C). A further 621 samples from farms, covering most of Sweden, were used to validate the results. Standard error of prediction (SEP) for metabolizable energy varied between 0.3-0.5 MJ kg DM-1 depending on botanical composition and conservation method of the forage. Energy values in 70% of the samples from farms were predicted equal to or better than 0.5 MJ kg DM-1 from the reference value. The range of SEP for crude protein was 5-7 g kg DM-1. On the samples from farms only 50% of these were predicted better than 7 g kg DM-1 due to a systematic difference obtained for one half1033of the samples. A new calibration on these samples (SEP=6 g kg DM-1) improved overall prediction to 63% below 7 g kg DM-1. It was concluded that provided the instrument and the calibrations are regularly checked against reference methods, NIRS analysis will meet the requirements for accuracy in assessing the feeding value of forages in practi
显示更多 [+] 显示较少 [-]