The aim of this study was to perform a NIR calibration model for crude protein prediction of forages harvested in June 2009 from hill permanent grassland (Gradinari, Caras-Severin District) organized in ten experimental trials fertilized organic, mineral, and organo-mineral. The soil was Calcic Luvisol and the annual average temperature around 10.4oC. The floristic composition was determined gravimetrically. From Poaceae were present Festuca rupicola and Calamagrostis epigejos. Fabaceae family

more »

... as represented by Trifolium repens and Lathyrus pratensis. From other botanical family: Rosa canina, Filipendula vulgaris, Galium verum and Inula britanica. Like input data was used the results for this parameter by Kjeldahl chemical method and the reflectance values from NIR spectra for analysed samples. Partial last square (PLS) regression was selected to perform the multivariate analysis to obtain the "NIR-CP" model, implemented in Panorama program (version 3, LabCognition, 2009). The statistical parameters (R2=0.8630; RMSEC=1.2844) and the differences between references and predicted values situated in range 0.03 - 1.73 % shows that it is promising to use this calibration model to evaluate the quality of forages from grassland in this period of year.