Probabilistic Cloud Masking for the Generation of CM SAF Cloud Climate Data Records from AVHRR and SEVIRI Sensors

Abstract

Cloud screening in satellite imagery is essential for enabling retrievals of atmospheric and surface properties. For climate data record (CDR) generation, cloud screening must be balanced, so both false cloud-free and false cloudy retrievals are minimized. Many methods used in recent CDRs show signs of clear-conservative cloud screening leading to overestimated cloudiness. This study presents a new cloud screening approach for Advanced Very-High-Resolution Radiometer (AVHRR) and Spinning Enhanced Visible and Infrared Imager (SEVIRI) imagery based on the Bayesian discrimination theory. The method is trained on high-quality cloud observations from the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) lidar onboard the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite. The method delivers results designed for optimally balanced cloud screening expressed as cloud probabilities together with information on for which clouds (minimum cloud optical thickness) the probabilities are valid. Cloud screening characteristics over 28 different Earth surface categories were estimated. Using independent CALIOP observations (including all observed clouds) in 2010 for validation, the total global hit rates for AVHRR data and the SEVIRI full disk were 82% and 85%, respectively. High-latitude oceans had the best performance, with a hit rate of approximately 93%. The results were compared to the CM SAF cLoud, Albedo, and surface RAdiation dataset from AVHRR data–second edition (CLARA-A2) CDR and showed general improvements over most global regions. Notably, the Kuipers’ Skill Score improved, verifying a more balanced cloud screening. The new method will be used to prepare the new CLARA-A3 and CLAAS-3 (CLoud property dAtAset using SEVIRI, Edition 3) CDRs in the EUMETSAT Climate Monitoring Satellite Application Facility (CM SAF) project.