(1) a set of albedo measures;
(2) suites of model parameters describing the anisotropy of the surface BRDF; and
(3) nadir BRDF-adjusted surface reflectances (NBAR).
These gridded products are in a sinusoidal tiled projection and HDF-EOS format and have the status Validated, Level 1. The algorithm providing BRDF and albedo fits a semiempirical BRDF model to MODIS observations on a pixel-by-pixel basis. When cloud-free looks are few in number for a pixel, the algorithm uses an adaptive strategy that exploits a BRDF database of prior knowledge.
Global land cover at 1-km spatial resolution is coded according to multiple land cover schemes:
IGBP; University of Maryland;
and Plant Functional Types.
Classification uses a database of MODIS land products assembled each month and taken for one year of observation. A supervised classification algorithm uses decision trees with boosting to estimate anterior probabilities of membership for each pixel in each class, then utilizes prior probabilities from ancillary data and prior land cover products to find posterior probabilities and assign class labels to pixels. The training database includes more than 1500 global training site exemplars assembled using Landsat data and collateral information as available for local regions. Global accuracy is in the range of 75-80 percent for the IGBP classification.
Copyright CSIRO 2003 ©
Use of this web site and information available from it is subject to our
Legal Notice and Disclaimer