Abstract
The utility of near infrared reflectance spectroscopy (NIRS) as a rapid nutritional evaluation tool is not yet fully defined for a number of non-conventional feedstuffs, which are considered to be of low economic value. This study was designed to calibrate and validate the NIRS technique as a potential tool to predict the nutritional value of browse leaves for ruminant animals. Acacia erioloba (E. Mey), A. nilotica (L. Delile) and Ziziphus mucronata (Willd) leaf samples were harvested, dried and milled (2 mm sieve) before being scanned (32 scans per spectra) from 1100 to 2500 nm with spectra being recorded at intervals of 2 nm using a SpectraStar XL. Spectral data were recorded in diffuse reflectance as log (1/R). The samples were then analysed for chemical composition, buffer-soluble nitrogen, and in vitro ruminal N degradation to generate reference values. Reference values for all samples were imported into the NIRS spectral data file and used to develop calibration equations. Data analysis and calibration were done with a UCal software (Unity Scientific, Australia). Calibration models were externally validated using reference values from an independent set of browse leaves. Total nitrogen (N) and acid detergent fibre (ADF) content showed good calibration statistics with high R2 values of 0.988 and 0.991; and standard error of calibration (SEC) of 0.452 and 13.628, respectively. Calibration models explained 42.8 % and 77.5% of the variation in N degradability at 24 hours of incubation (ND24) and at 36 hours (ND36). External validation showed that calibration models were able to predict the total N content of independent samples with R^2 and standard error of prediction (SEP) values of 0.82 and 18.19, respectively. However, for all the other nutritional parameters, validation statistics were poor with very low R^2 and high SEP values. It was, therefore, concluded that the NIRS models developed in this study can be used to accurately determine total N in browse leaves but not in vitro ruminal N degradability.
