An automated cloud detection method for daily NOAA 16 advanced very high resolution radiometer data over Texas and Mexico

Abstract

The advanced very high resolution radiometer (AVHRR) data acquired from the National Oceanic and Atmospheric Administration (NOAA) satellites have been widely applied to a variety of environmental research. A single AVHRR scene is seldom completely cloud‐free. Maximum value compositing (MVC) to create a single image from multiple orbits and dates has become the most valuable method to minimize cloud contamination. Composite images are not absolutely cloud‐free. Postcomposite cloud screening of the composite aggregates was developed to overcome the residual cloud contamination problem, but this is not possible for real‐time delivery of composite data or not suitable for compositing based on AVHRR data from multiple NOAA satellites. Another approach is to detect and remove cloud‐contaminated pixels from daily AVHRR scenes prior to applying the MVC method to provide real‐time composite images. This study developed an automated cloud detection method for daily NOAA 16 AVHRR scenes over the state of Texas and Mexico. The accuracy of the cloud detection algorithm was greater than 93% based on a random test sample from 36 images. Unidentified cloud shadow pixels as well as misidentified barren land pixels and water pixels contributed to more than 5% of the accumulated errors. The error from misidentification of water pixels can be reduced by assigning different threshold values for channel 4 brightness temperature according to the geographical latitude of the data. The resulting daily cloud‐free AVHRR data can be used to construct short‐time period composite images valuable for detecting subtle but critical environment changes. In addition, compositing methods other than MVC, such as multidate averaging or minimum value selection, can be applied for various research purposes, once the daily AVHRR data are cloud‐free.