Accuracy of single and mix calibration as method to determine ruminant feedstuffs nutrient utilization using Near-Infrared Reflectance Spectroscopy

Abstract

Including real-time data nutrient utilization from quick and accurate methods such as NIRS in a ration formulation improves feed efficiency. However, the homogeneity of the feedstuff sample set used in calibration influences its prediction accuracy and application. This study compares the calibration accuracy of ruminant feed nutrient utilization using a single or mixed calibration. Feedstuffs of different origins (Padang (24), Bandung (24), and Bogor (24)) consisting of forages (30) and concentrate (30) were used separately for single calibration or together for mixed calibration. 12 samples were used for external calibration. The spectrum of each sample was collected using FT-NIRS. The calibration used the PLS regression model. Parallel to the NIRS spectrum collection, nutrient utilization was observed using in vitro methods including degradation of dry matter, organic matter, crude protein, rumen pH, VFA, and NH3 concentration. The results showed accuracy of nutrient utilization prediction varied between the parameters observed. Except for VFA, mixed calibration produced better accuracy (R2>0.83 and RPD>2.21) than single calibration. External calibration proved that NIRS prediction was not significantly different from the in vitro results (P > 0.05). It is concluded mixed ruminant feedstuffs database can be used accurately and broadly to estimate nutrient utilization using NIRS.