Most previous species classification efforts have been limited to a
small number of sites. Roth et al. (2015-in this issue) compare multiple
species classification approaches in five different North American
ecosystems. Classification accuracies varied by site and vegetation
functional type, with maximum accuracies for each site ranging between
61% and 92%. Linear discriminant analysis provides higher
average accuracy compared to multiple endmember spectral mixture
analysis (MESMA), and dimension reduction techniques show promise
for maintaining or increasing classification accuracy. Since HyspIRI is a
global mapping mission, the work described in Roth et al. (2015-in
this issue) is critical for determining which techniques are likely to be
useful for mapping species across larger scales