A PIXEL-SCALE ALGORITHM OF CLOUD TYPE, LAYER, AND AMOUNT FOR AVHRR DATA .1. NIGHTTIME

An automated pixel-scale algorithm has been developed to retrieve clou d type, related cloud layer(s), and the fractional coverages for all c loud layers in each AVHRR (Advanced Very High Resolution Radiometer) p ixel at night. In the algorithm, cloud-contaminated pixels are separat ed from cloud-free pixels and grouped into three generic cloud types. Cloud layers in each cloud type are obtained through a cloud-type unif ormity check, a thermal uniformity check, and a channel 4 (11 mu m) br ightness temperature histogram analysis, within a grid area. The algor ithm allows for pixels to be mixed among different cloud layers of dif ferent cloud types, as well as between cloud layers and the ocean or l and surface. A ''neighbor-check'' method is developed to identify the cloud layers associated with each mixed pixel and to calculate the cov erages of each of the cloud layers in the pixel. Digital color images are generated based on information on the location, cloud type, cloud layer, and cloud amount of each individual pixel. Visualization compar isons show good agreement between color-coded images and the standard black and white satellite images. The results of the pixel scale algor ithm also show good agreements with the spatial coherence analysis and with National Weather Service surface and radiosonde observations. Th e pixel-scale algorithm has been developed for use in validation of ou tput from CLAVR (clouds from AVHRR) project algorithms.