Prediction of gas production kinetic parameters of forages by chemical composition and near infrared reflectance spectroscopy

This study was initiated to evaluate the potential of near infrared reflectance (NIR) spectroscopy to predict in vitro gas production parameters of botanically complex herbage samples. A total of 94 herbage samples harvested in natural meadows located in the mountains near León in Northwest Spain were analyzed to determine their chemical composition. In addition, all herbage samples were incubated in vitro in buffered rumen fluid to determine fermentation kinetics using a gas production technique, and scanned in a spectrophotometer to obtain NIR spectra. Prediction equations showed that NIR spectra could explain a high proportion of the variability (R2 > 0.94) related to some in vitro gas production parameters (e.g., fractional rate of fermentation (c) and extent of degradation in the rumen at different passage rates (ED03 and ED06)) of the calibration set (n = 62). When these NIR equations were applied to the validation set (n = 32), most parameters were satisfactorily predicted with standard errors of prediction (SEP) of 3.88 ml for gas production at 24 h of incubation (GP24), 2.71 ml for asymptotic gas production (A), 0.0038 for c, 0.020 for ED03 and 0.018 for ED06, accounting for less than 7% of the corresponding mean value. However, lag time (L) could not be predicted by NIR spectroscopy. The SEP was always lower when NIR spectra were used as predictors in comparison with chemical composition, perhaps because spectra contained information about feed constituents as well as physical properties of the samples. Nevertheless, results suggest the need for improved standardisation of this gas production procedure, to minimise the influence of sources of experimental error to increase repeatability and reproducibility, in order to obtain accurate NIR determination of feed fermentation kinetics and of parameter estimates.