TOWARDS BETTER QUALITY OF AVHRR COMPOSITE IMAGES OVER LAND - REDUCTION OF CLOUD CONTAMINATION

A procedure for reducing cloud contamination in NOAA/AVHRR composite imagery is proposed and applied to the NOAA global vegetation index data set. The suggested approach is based on thresholding the AVHRR/Channel 4 (11 mu m) brightness temperature. The global spacetime-angle dependent thresholds (monthly, 2 degrees x2 degrees, for 30 degrees-viewing angle bins) are estimated in two steps. First, a clear-sky background for Channel 4, in terms of means and standard deviations, is developed from GVI weekly composite data using their ''greenest'' subsamples. Second, this background is used to estimate the space-time-angle dependent thresholds. Both the size of the ''greenest'' subsample for deriving the climatology and its use for constructing the thresholds are discussed in detail using targets in central United States and Thailand. The approach is generalized further and applied globally. Special attention is devoted to the development of quantitative criteria to estimate the efficiency of both steps and to the improvement resulting from implementation of the proposed procedure of cloud screening.