MAPPING DESERT SHRUB RANGELAND USING SPECTRAL UNMIXING AND MODELING SPECTRAL MIXTURES WITH TM DATA

Spectral unmixing experiments were done to explore the applicability o f linear unmixing models, especially the basic least-squares method fo r mapping sparse vegetation in rangeland. Some important theoretical a nd technical issues involved in physical inversion problems were addre ssed. Based on the field reference spectra of image components, a cons trained least-squares method was applied to Landsat Thematic Mapper da ta over an area in Long Valley, Nevada to calculate vegetation abundan ce in a pixel. A method for formulating a well-conditioned spectral mi xture by calculating the cosine of the angles between the candidate su rface components was presented. This method provides a way to measure the separability of candidate endmembers quantitatively and derive spe ctral endmembers objectively. The results of this study suggest that t he ambiguity or uncertainty in physical inversion problems arises from the inability to provide a complete set of representative reference s pectra and to formulate a well-conditioned spectral mixture, not from the least-squares method itself. Some of import implications of the st udy include the following: (1) the unmixing techniques can provide mod erate estimates of vegetation fractions in arid rangeland, where veget ation is sparse, with TM data; and (2) the degree of spectral pureness of endmembers should be consistent between endmember spectra that are used for unmixing.