SEVIRI Cloud mask by Cumulative Discriminant Analysis

In the context of cloud detection for satellite observations we want to use the method of Cumulative Discriminant Analysis (CDA) as a tool to distinguish between clear and cloudy sky applied to Spinning Enhanced Visible and Infrared Imager (SEVIRI) data. The methodology is based on the choice of several statistics related to the cloud properties, whose correlation has been analyzed by Principal Component Analysis (PCA). Results have been compared with the SEVIRI reference cloud mask provided by the European Centre for the Exploitation of Meteorological Satellite (EUMETSAT), in order to find suitable thresholds able to discriminate between clear or cloudy conditions. We trained the statistics on a selected region, the Basilicata area, located in the south of Italy, in different periods of the year 2012, in order to take into account the seasonal variability. Moreover we separated land and sea surface and distinguished between day-time or night-time. The validation of thresholds, obtained through SEVIRI observations analysis, shows a good agreement with the reference cloud mask.