Cloud detection for the Far East region using NOAA AVHRR images

A new automatic cloud detection method to process the large data sets required for long-term trends in AVHRR time series images is described and tested over the Far East region. In the cloud detection method, a simple NDVI test was used to roughly separate land from ocean. After the NDVI test we carried out two tests: a visible or near-infrared test and an infrared test. Results from the cloud detection method over the Far East region are presented. In order to assess the cloud detection method, a comparison was carried out between the method presented and the Saunders and Kriebel daytime procedure using the N-Land database. We found that in general, the method presented has better performance than the Saunders and Kriebel procedure by about 0.5% to 10.5%.