The distribution of errors associated with genotype and environment during the prediction of the water-soluble carbohydrate concentration of perennial ryegrass cultivars using near infrared reflectance spectroscopy

Abstract

Significant deviations associated with site or cultivars within sites were detected in 4 of 6 independent near infrared reflectance spectroscopy (NIRS) calibrations developed to predict water-soluble carbohydrate (WSC) concentrations in perennial ryegrass herbage harvested from 2 sites. These effects were observed both when calibration subsets were selected on the basis of spectral characteristics, and when calibration sets were balanced with respect to a priori knowledge of the structure of the data set. However, there were also instances when non-random deviations were not significant, demonstrating that it was possible to develop broadly based NIRS calibrations to predict WSC in perennial ryegrass. Deviations between NIRS predictions and reference values should be monitored, with reference to the structure of the experiment from which the samples were derived, before NIRS estimates of WSC concentration are used in agronomy or plant breeding.