Abstract
The objective was to evaluate the adequacy of near-infrared spectroscopy for predicting the chemical-bromatological composition of sorghum BRS 658 and BRS Ponta Negra silage, in PVC mini-silos, ensiled at grain maturation stages E7 or milky, E8 or pasty and hard mealy. The model using the Savitzky-Golay 1st Derivative for DIVMS showed lower RMSEC, RMSECV and RMSEP (5.96, 7.42, 6.96) respectively. For DIVMO, the Multiplicative Signal Correction (MSC) + Savitzky-Golay 1st Derivative model with two windows showed better performance, with higher R2 in calibration, validation and prediction (0.82, 0.72, 0.56) respectively. In the independent validation of the models, no significant difference was observed between the predicted and reference values for the BP, IVMSD and IVMOD variables. The NIRS methodology is capable of accurately predicting the protein content, in vitro digestibility of dry matter and in vitro digestibility of organic matter in sorghum silage.