Employing ground-based spectroscopy for tree-species differentiation in the Gulf Islands National Park Reserve

Airborne hyperspectral data is a promising tool to map species distribution; however, the large number of input bands can be highly correlated and potentially noisy. Ground-based spectrometer data can identify spectral regions that are optimal for species differentiation, and therefore provide a logical initial step for species mapping endeavours employing airborne hyperspectral data. This study used reflectance collected by an Analytical Spectral Devices (ASD) spectrometer to differentiate between tree species common to the Canadian Gulf Islands. Baseline ASD reflectance and its derivatives were used as input for forward stepwise discriminant analyses to identify wavelengths that minimize within-species variance while maximizing between-species variance. Identified wavelengths were then used as input for normal discriminant analyses, which confirmed through cross-validation classifications that, at the leaf scale, species could be differentiated with an overall accuracy. 98% and individual accuracies. 85% using 40 optimal wavelengths. Accuracies slightly decreased when using derivatives, but only for certain species. Results indicate that wavelengths in the ranges 501-550, 681-740 and 1401-1800 nm exhibited the most significance. The selected bands form the basis of ongoing mapping efforts using airborne hyperspectral imagery.