Extending near infrared reflectance(NIR) pulp yield calibrations to new sites and species

Recently, we demonstrated that the accuracy of pulp yield predictions for wood samples from a site (Gog) new to their calibration (Tasmania-wide Eucalyptus nitens) was greatly improved by adding five Gog samples to the calibration set. In this study we investigated the addition of Gog samples to the Tasmania-wide E. nitens set, with the aim of further improving predictive accuracy. It was demonstrated that the addition of a single Gog sample to the Tasmania-wide calibration set was sufficient to greatly reduce predictive errors and that the inclusion of at least 3 Gog samples in the Tasmania-wide set was sufficient to give relatively stable predictive errors. The addition of different sets of 5 Gog samples to the Tasmania-wide calibration, however, caused predictive errors to vary between sets. The standard deviation of pulp yield for the prediction set (20 Gog samples) was important, with sets having the largest standard deviations giving the best predictive statistics. Finally, the Tasmania-wide E. nitens calibration was enhanced using samples from a different species (Eucalyptus globulus) and applied successfully to other E. globulus samples.