Comparison of green leaf eucalypt spectra using spectral decomposition

Abstract

Vegetation function and dynamics are key parameters in terrestrial carbon-cycle models. The strong linkages between biochemical constituents in foliage with photosynthetic capacity and ecosystem productivity makes the development of methods to characterise patterns of foliage biochemistry a potentially powerful approach for estimating leaf function and carbon fluxes at a variety of scales. Eucalypt foliage spectra were obtained over a range of species and locations in southern New South Wales, covering a significant productivity and climatic gradient. We applied a spectral decomposition technique, based on multivariate factor analysis, which allows inter-correlations of underlying factors affecting a set of variables to be assessed. A small number of factors capture virtually all of the variation observed in the foliage spectra and each factor contains significant information relating to species and plot variation over the region. Factor analysis indicated that key chlorophyll, nitrogen, protein and water absorption features could be accurately identified across the spectra. In addition, significant correlations existed between factor loadings and environmental data of the region, including mean annual rainfall and a soil fertility index.