Abstract
Near-infrared reflectance spectroscopy (NIRS) was used to predict chemical composition of unground (cut to 5cm length) forage paddy rice samples (n=120). Eighty samples were selected for calibration development, and the remaining samples were used for equation validation. Calibration equation for dry matter (DM), crude protein (CP), organic cellular contents (OCC), organic cell wall (OCW) and low digestible fiber (Ob) showed a tendency for increasing correlation coefficients (r) between estimated and measured values and decreasing standard error (SE) of estimation as the number of spectra (repetitions) used for equation development increased, with the most accurate calibration equations being produced by averaging 4 or 5 repetition spectra. The validation test showed relatively high values of r and low values of standard deviation of prediction (SDP) when equations developed from 2-5 spectra were used for DM and those from 4-5 spectra were used for CP, OCC, OCW and Ob. The results show that DM, CP, OCC, OCW and Ob contents in unground forage paddy rice can be predicted by using a NIRS technique with an acceptable accuracy, when sufficient number (4-5) of repetitive spectra is incorporated into the equation development.
