The use of MODIS reflectance anisotropy to recover land surface properties

This study explores the use of reflectance anisotropy as described by the Bidirectional Reflectance Distribution Function (BRDF) to recover land surface properties. The effort primarily utilizes the reprocessed V005 MODerate Resolution Imaging Spectroradiometer (MODIS) BRDF/Albedo products, which have been produced at a 500 m resolution every 16 days since 2000. One measure of reflectance anisotropy is the Anisotropie Flat Index (AFX) that is defined by the ratio of white sky albedo (WSA) to the isotropic parameters (ISO) of the RossThick-LiSparse-Reciprocal (RTLSR) BRDF model. An investigation of the AFX demonstrates that this BRDF measure captures surface anisotropic patterns that are related to vegetation structure. Two regional case studies of the Canadian boreal forest and the Australian savanna determine the ability of AFX and various other BRDF measures to improve the land cover classification accuracies produced by a decision tree classifier (c4.5).;AFX, geometric and volumetric parameters, and several other BRDF shape indicators are all derived from the semi-empirical kernel-driven BRDF model that is routinely produced by the MODIS BRDF/Albedo product. These BRDF measures are evaluated for their potential as an additional source of information in addition to the spectral signatures that are the conventional inputs to land cover classifiers.;This research indicates that the inclusion of BRDF features can significantly reduce the confusion among those classes with canopy structural variations that are difficult to discern with remotely sensed spectral reflectance signatures alone. An approximately 5 percent improvement in overall accuracies are achieved by including BRDF features in both these case studies. The greatest improvements are seen for the boreal Wetland Shrub class with user and producer's accuracies increasing by 17.7 and 11.3 percent, and for the Australian Eucalyptus miniata woodland with grassland understory class with user and producer's accuracies increasing by 11.65 and 18.33 percent. These findings indicate that the inclusion of reflectance anisotropy information can contribute to an improvement in the discrimination of global land covers which serve as important initial conditions in climate and biogeochemical models.