Unmixing the directional reflectances of AVHRR sub-pixel landcovers

Recent progress in canopy bidirectional reflectance distribution function (BRDF) model inversions has allowed accurate estimates of vegetation biophysical characteristics from remotely sensed multi-angle optical data. Since most current BRDF inversion methods utilize one-dimensional (1D) models, surface homogeneity within an image pixel is implied. The Advanced Very High Resolution Radiometer (AVHRR) is one of the few spaceborne sensors capable of acquiring radiometric data over the range of view angles required for BRDF inversions. However, its relatively coarse spatial resolution often results in measurements of mixed landcovers, and thus the data may not be ideal for BRDF inversions. The authors present a three-step spectral unmixing method for retrieving AVHRR sub-pixel directional reflectances in regions of high spatial heterogeneity. The reflectances of individual vegetation types are deconvolved using co-located Landsat TM and AVHRR data. The three major steps in the model include: (1) unmixing of vegetation endmember concentrations in TM imagery; (2) correction of dissimilar shadow fractions between TM and AVHRR data; and (3) unmixing of AVHRR sub-pixel reflectances of vegetation types for any Sun-sensor geometry. The authors tested the method using simulated TM and AVHRR data. A savannah landscape simulation, comprised of a canopy radiative transfer model and a crown geometric-optical model was used to create images containing mixed pixels of tree, grass, and shade endmembers. TM and AVHRR spectral response functions, viewing geometrics, and off-nadir pixel shape calculations were incorporated into the simulations. Following the successful testing of the unmixing method on error-free simulations, random noise representing atmospheric perturbations and co-registration inaccuracies was added to the data. The method is stable when errors resulting from either the first unmixing step or image co-registration inaccuracies are introduced. Potential errors in the AVHRR data may result in inaccurately retrieved reflectances if the image scene contains a spatially homogeneous mix of landcovers. A method for detecting and mitigating this problem is presented