Cloud detection with GOME: a refinement of the cloud clearing algorithm using ATSR-2 images

The Global Ozone Monitoring Experiment (GOME), flown on-board the ERS2 satellite since 1995, has the main mission of retrieving total ozone at the nominal ground resolution of 320×40 km². The retrieval of different trace gases and aerosol can also be attempted. Cloud detection and characterization, an interesting result in itself, are needed to analyse spectral data prior the retrieval of columnar ozone as well as other atmospheric constituents. The Polarisation Measurement Devices (PMDs) allow for detection of radiation leaving the Earth-atmosphere system at three spectral broad channels, from 300 to 800 nm. The Cloud Clearing Algorithm (CCA) was developed based on a simple thresholding method: cloud detection is obtained within the PMD ground pixel (20×40 km², one-sixteenth of GOME´s spatial resolution) using thresholds; that depend primarily on surface type and reflection, and solar zenith angles. A refinement of the CCA is presented hereafter. Thresholds over the ocean have been computed by comparing PMD detection performances with the Along Track Scanning Radiometer 2 (ATSR2) cloud masks. ATSR2 masks are available on a 2×2 km² spatial resolution. Note that GOME and ATSR2 do fly on-board the same spacecraft, thus producing simultaneous nadir images with very reliable colocation. Refined CCA performances have been compared with a totally independent cloud classification algorithm that uses visible-infrared, high resolution full disk METEOSAT images. Case studies are presented, and differences between the two methods are discussed at PMD and spectral GOME ground pixel sizes