PREDICTION OF THE FEEDING VALUE OF MAIZE SILAGES BY CHEMICAL-PARAMETERS, IN-VITRO DIGESTIBILITY AND NIRS

Fifty maize silages with known in vivo organic matter digestibility (V OMD) were used to compare the potential of chemical parameters (dry ma tter (DM), crude protein (CP), crude fibre (CF), crude fat (Cfat), ash , neutral detergent fibre (NDF), acid detergent fibre (ADF), acid dete rgent lignin (ADL), starch (STA), in vitro digestibility (with rumen f luid (RFOMD) or commercial cellulases (COMD)) and NLRS (monochromator: 1100-2500 nm) in predicting VOMD and calculated metabolizable (ME) an d net energy lactation (NEL). A second series of 51 maize silages with digestibility and energy value based on RFOMD served as test set. Fur ther, the possibilities of NIRS to predict chemical composition and ca lculated protein value (DVE and OEB) were investigated. NIRS-equations , developed by partial least squares analysis, were tested by internal and external validation. The correlation with VOMD was highest for CO MD (r = 0.82), followed by RFOMD (0.81), ADF (-0.77) and NIRS-predicte d VOMD (0.75). ADL could explain some supplemental variance in VOMD be sides ADF and RFOMD, but not with COMD. Multiple linear regressions to predict ME and NEL, irrespective of being based on chemical parameter s, RFOMD or COMD had similar residual standard deviations of about 0.2 1 and 0.16 MJ kg(-1) DM, respectively. With NIRS the standard error of prediction amounted to 0.28 and 0.20 MJ kg(-1) DM, respectively. The relative ability of NIRS to predict chemical composition was very good for residual moisture, STA, ADF, NDF and CP, moderate for CF and ADL and low for ash. The prediction potential for OEB was moderate and nih il for the little varying DVE-content. (C) 1997 Elsevier Science B.V.