INVESTIGATION OF LEAF BIOCHEMISTRY BY HIERARCHICAL FOREGROUND BACKGROUND ANALYSIS/

A hierarchical procedure was developed for quantitative estimation of foliar chemistry from remote reflectance spectra, We based our analysi s on a new methodology called Hierarchical Foreground and Background A nalysis (HFBA) that derives sequentially a series of weighting vectors which simultaneously extract important discriminant features, in this case, leaf anatomy and chemical concentration at different levels of detection from the spectral information. In this study, we focused on the application of detecting carbon, cellulose, nitrogen concentration s, and water content. The goal of the derived vectors is twofold: 1) c reate a robust detection and classification system of constituent mate rials and 2) create a good information packing system that minimizes e xtraneous undesired interference, like noise, in the analysis, In our study, two data sets were examined: a fresh leaf (FL) data set, LOPEx [from the Joint Research Center (JRC), Ispra, Italy], and a dry leaf d ata set, Blackhawk Island (BH), WI [University of New Hampshire (UNH), Durham], We tested the robustness of the derived vectors with four ot her data sets: fresh leaf data from Jasper Ridge Biological Preserve ( JRBP) (chemistry from UNH, spectra from University of California, Davi s), Santa Monica Mountains (SMM), CA, [from University of California, Davis (UCDavis)], and dry leaf data from two ACCP sites (UNH); Howland (HO), ME, and Harvard Forest (HF), MA. The results support the robust ness of the HFBA system and demonstrate an advantage in classification accuracy (first level) as well as in predicting the biochemical compo sition (subsequent levels) over classical forms of analysis that ignor e effects of the nonlinear variation that contribute to reflectance at different (subpixel and spectral) scales. HFBA primarily deals with t he spectral scaling issue.