Determination of Forage Quality by near Infrared Reflectance Spectroscopy: Efficacy of Broad-Based Calibration Equations

Abstract

Two experiments were conducted to determine 1) if statistical clustering of near infrared spectra would aid in selection of samples to establish calibration equations, and 2) if broad-based calibration equations were capable of accurately determining forage quality. In Experiment 1, clustering of spectra did not have any advantage over random selection as a means to select samples. In Experiment 2, 990 hay samples representing a large diversity of species, maturity, cutting, and chemical composition, were collected from 31 states. Approximately 50% of the samples were used in this study. Samples were separated into calibration and validation sets, either on a random basis or by subset of samples from individual states into validation sets. Based on randomly selected samples, the standard error of validation and bias were dry matter (.47, -.05%); CP (.84, .06% dry matter); ADF (2.24, .18% dry matter); NDF (2.16, 17% dry matter); in vitro dry matter digestibility (30.3, -.40%). There was a trend toward increased bias for ADF, NDF, and in vitro dry matter digestibility when samples from particular states, rather than randomly selected samples, were used as validation sets.