Broad NIRS calibrations to predict nutritional value of the southern feedbase

Near infrared reflectance spectroscopy (NIRS) is used to predict the nutritive characteristics of feeds consumed by livestock. Once calibration equations are developed between reflectance and measured nutritional traits, NIRS is rapid and inexpensive. Our aim was to develop broad NIRS calibrations for the southern feedbase of Australia. A total of 4385 samples from 154 accessions of 109 species of annual and perennial legumes, grasses and forbs were grown in common plots at two locations over 3 seasons. Plots were sampled across all growth stages. A quarter of these samples were subject to laboratory analysis of dry matter digestibility (DMD), total nitrogen (N), neutral detergent fibre (NDF), acid detergent fibre (ADF) and organic matter (OM). Of the samples analysed, half were used to develop the calibration and half were set aside for validation. Development of broad calibrations across the sample range was very successful despite the large variation in the taxonomy and life history of the samples. When predicting samples from the collection that were not included in equation development, the statistics of prediction were; total N - r 2 0.96, RPD (relative percent difference) 5.3, in vitro DMD - r 2 0.93, RPD 3.7, ADF - r 2 0.93, RPD 3.9 and NDF - r 2 0.95, RPD 4.3. When the validation samples were separated into taxonomic groups, the prediction errors were considerably lower for annual species than perennial species. The data generated is being used to compare nutritional value of species over time and investigate opportunities to improve productivity and reduce methane emissions intensity from sheep in southern Australia.