Rapid discrimination of eucalypt species using a handheld near-infrared instrument

Abstract

Eucalypt trees are a key feature of many Australian ecosystems; however, it can be difficult to identify the various species based on their leaves alone. The rapid, non-destructive analytical technique of near-infrared spectroscopy (NIRS) has shown promise for discrimination of similar species in other flora types; thus this study investigated the prospect of using NIRS to differentiate between the leaves of six eucalypt species from two genera (Corymbia and Eucalyptus). The results showed good results for the discrimination of genera using PLS-DA (98 % accuracy) and SIMCA (72–90 % accuracy). Moving window analysis showed that the region between 1300 and 1500 nm – which includes CH and OH bonds – had the largest influence on model accuracy. Discrimination of the six species was more challenging, with PLS-DA showing accuracies of 83–97 % for cross-validation. However, SIMCA showed more acceptable results, with accuracies of 100 % for cross-validation and 79–90 % for a dependent test set. These results suggest that NIRS combined with appropriate chemometric methods may be suitable for the rapid discrimination of leaves from different eucalypt genera and possibly different species.