Prediction of in situ degradation characteristics of neutral detergent fibre (aNDF) in temperate grasses and red clover using near-infrared reflectance spectroscopy (NIRS)

Near-infrared reflectance spectroscopy (NIRS) was used to estimate in situ neutral detergent fibre (aNDF) degradation characteristics. Dried or ensiled samples of perennial grasses and red clover (n = 382) were used to develop calibration equations. In situ residues of aNDF after 144 h of ruminal incubation were assumed to represent the total indigestible aNDF fraction (aINDF). The fractional degradation rate of aNDF (kdNDF) was calculated assuming first order degradation kinetics with an initial lag time. The aNDF content ranged from 266 to 703 g/kg DM. The aINDF values ranged from 14 to 360 g/kg DM, and the kdNDF varied from 0.020 to 0.137/h. Spectral data was transformed to their second order derivates and a standard normal variate and detrending procedure was used. NIRS prediction were made by modified partial least square technique and prediction accuracy was tested after separating the samples into a validation (close to 70 samples) and calibration dataset. Establishment of new NIRS calibrations and successive predictions of new validation samples were repeated four times. Accuracy of NIRS predictions was evaluated by the standard error of prediction, corrected for bias (SEPC), and by the standard coefficient of determination in prediction (RP2). Dispersion in sample reference values (square root of variance) divided by SEPC (RPD) was also calculated and a value >2.5 was used as indicator of acceptable prediction. For aINDF, SEPC was 17.7 g/kg DM, Rp2 was 0.92 and RPD 3.4. For kdNDF, SEPC was 0.012/h, Rp2 was 0.51, and the RPD value was 1.7. For the initial lag time, SEPC was 1.27 h, Rp2 was 0.02 and RPD was 1.0.From this study it is concluded that the developed calibration equations had an acceptable prediction of aINDF, and can be used for predicting aINDF for use in practical feed evaluation. The NIRS prediction of kdNDF was less accurate because of weaknesses with the reference method. The initial lag time could not be predicted by NIRS.