An automated cloud detection method for daily NOAA-14 AVHRR data for Texas, USA

Publisher Website
Google Scholar
Abstract
A variety of cloud types appears in each Advanced Very High Resolution Radiometer (AVHRR) image. Clouds may contaminate solar reflectance data to be used for vegetation studies. This may jeopardize the accuracy of any quantitative results from data analysis. Published cloud detection algorithms for AVHRR data to date have mainly used data over Europe received from the National Oceanic and Atmospheric Administration (NOAA)-12 or earlier satellites. This study examined the previously published cloud detection methods with the intent to develop an automated cloud detection algorithm for NOAA-14 AVHRR data for Texas. Through testing a whole year of AVHRR scenes, the Texas automated cloud detection algorithm was capable of correctly identifying most of the cloudcontaminated pixels except for cloud shadow pixels. The overall accuracy reached 89%. The developed algorithm includes three major steps, top-of-the-atmosphere reflectance of channel 1, temperature difference of channels 3 and 4, and a combination of ratio of channel 2 to channel 1 and temperature in channel 4.