Characterizing the spatial variability of broadband albedo in a semidesertenvironment for MODIS validation

Global data sets on land surface albedo will be one of the core products to be derived from data acquired by the Moderate Resolution Imaging Spectrora diometer (MODIS), part of NASA's Earth Observing System. Widespread accepta nce of this product by the user communities is dependent, in part, on a com prehensive and rigorous programme of calibration and validation. Since the MODIS albedo product will be produced at a spatial resolution of 1 km, whil e measurements obtained from field instruments typically relate to areas of only a few tens of meters, this requires an understanding of the spatial v ariability of land surface albedo and a robust means of scaling zip from fi eld to satellite measurements. In this article, we examine these issues for a semidesert environment (the PROVE'97 field site at Jornada, New Mexico, USA). Spatial variations in field measurements of broadband albedo are rela ted to the functional ground cover of different scene elements (live and se nescent vegetation, soil and shadow) via a simple linear mixture model. Inf ormation on the fractional ground cover of the scene elements is derived fr om ground-based hemispherical photography. It is shown that the albedo valu es predicted by the mixture model are accurate to within 2% of the correspo nding measured values. This approach offers considerable potential for the validation of MODIS-derived albedo values through the use of spectral mixtu re modelling applied to fine spatial resolution satellite sensor images. (C ) Elsevier Science Inc., 2000.