Endmember bundles: A new approach to incorporating endmember variability into spectral mixture analysis

Accuracy of vegeation cover fractions, computed with spectral mixture analy sis, may be compromised by variation in canopy structure and biochemistry w hen a single endmember spectrum represents top-of-canopy reflectance. In th is article, endmember variability is incorporated into mixture analysis by representing each endmember by a set or bundle of spectra, each of which co uld reasonably be the reflectance of an instance of the endmember. Endmembe r bundles are constructed from the data itself by an extension to a previou sly described method of manually deriving endmembers from remotely sensed d ata. Applied to remotely sensed images, bundle unmixing produces maximum an d minimum fraction images bounding the correct cover fractions and specifyi ng error due to endmember variability. In this article, endmember bundles and bounding fraction images were create d for an airborne visible/infrared imaging spectrometer (AVIRIS) subscene s imulated with a canopy radiative transfer/geometric-optical model. Variatio n in endmember reflectance was achieved using ranges of parameter values in cluding leaf area index (LAI) and tissue optical properties observed in a N orth Texas savanna. The subscene's spatial pattern was based on a 1992 Land sat Thematic Mapper image of the study region, Bounding fraction images bra cketed the cover fractions of the simulated data for 98% of the pixels for soil, 97% for senescent grass, and 93% for trees. Averages of bounding imag es estimated fractional coverage used in the simulation with an average err or of less than or equal to 0.05, a significant improvement over previous m ethods with important implications for regional-scale research on vegetatio n extent and dynamics.