Spectral–age interactions in managed, even‐aged Eucalyptus plantations: application of discriminant analysis and classification and regression trees approaches to hyperspectral data

Abstract

The objective of this research was to evaluate species discrimination and specific age interactions in managed, homogeneous, even‐aged Eucalyptus and Acacia plantations (4–11 years old) in KwaZulu‐Natal, South Africa using hyperspectral data. CASI (36 bands; 426–952 nm spectral range; 1 m spatial resolution) radiance data from the spring (October) of 2004 were used to (i) discriminate between Eucalyptus and Acacia species and (ii) differentiate among age classes of the same species. Description of the nature of differences among various species and the spectral variability found in monoculture, even‐aged forest stands were of particular interest. Implications of this research relate to the use and interpretation of accepted forest state indicators, and the possible effect of age–site interactions on indicator outcomes. A secondary objective was to compare stepwise discriminant (variable reduction) analysis, followed by discriminant classification, to a classification and regression tree (CART) approach for both radiance and transformed minimum noise fraction (MNF) data. Results for discrimination accuracies using (MNF) transformed data were as high as 85% and 97% for between species and between‐species‐within‐age group (7–11 years) classifications, respectively. These results underlined the potential of spectral data of reduced dimensionality to be used in a Eucalyptus forest inventory context.