Pixel unmixing for hyperspectral measurement of foliar chemistry in Pacific Northwest coastal forests

Studies for the detection and mapping of variation in foliar chemistry have concentrated on the correlation of wavelength specific reflection and concentrations of foliar pigments and nutrients either through ground-based radiometric measurements or airborne data. The advantage of the former is that the scene components can be effectively controlled so that a relatively simple reflectance model can be constructed and end members extracted. In the case of using airborne data, however, the influence of scene components that mask, or subdue, the reflectance-chemical signal, may dominate. This has led to the development of methodologies for which the various scene components can readily be isolated and accounted. Pixel unmixing to isolate canopy reflectance from other scene components has long been used in the assessment of foliar characteristics and processes. Unfortunately the traditional methods of unmixing rely on distinct spectral signatures from the various scene components. This paper details a method developed to isolate the scene components when they are not spectrally dissimilar