Cloud fraction determination for Sentinel-4, -5 and -5P

Kurzfassung

An important step during the retrieval of atmospheric trace gases is the determination of the presence or absence of clouds and the characterization of their extent, usually expressed as a cloud fraction, because clouds have a significant influence on the accuracy of the trace gas retrieval. In contrast to many cloud detection algorithms working on sounders or on imagers in the infra-red or micrometer range, the Optical Cloud Recognition Algorithm (OCRA) is developed for sensors operating in the UV/VIS part of the spectrum, as covered by e.g. Sentinel-4, -5 or -5P. By design, OCRA can be applied to both, broad band and narrow optical wavelength ranges. Due to its RGB color space approach, a basic requirement for OCRA to be applied is that there is information available from the blue, green and red parts of the optical spectral range. The basic idea behind OCRA is to split the measurement of a scene into contributions of clouds and a cloud-free background, i.e. the reflectance in the absence of clouds. OCRA uses the general assumption that clouds have a higher reflectivity than the surrounding ground in all optical wavelengths. In the optical part, the cloud reflectivity is almost wavelength-independent and therefore clouds appear white in the normalized RGB color space since all three components are represented with the same amount. The scene which is furthest away from the white situation is the one where we expect the least possible amount of cloud contamination. Joining all these scenes on a global grid then provides the cloud-free background map, which is usually based on several years of data. The comparison of a measured reflectance with the corresponding predetermined cloud free reflectance can then be used to derive a radiometric cloud fraction for this given scene. With respect to earlier versions, OCRA now also includes degradation corrections for the reflectances as well as corrections for viewing zenith angle dependencies and latitudinal and seasonal dependencies. OCRA is well established for polar orbiting sensors like GOME, SCIAMACHY and GOME-2. An adaptation of OCRA to the polar orbiting Sentinel-5 and -5P is more or less straightforward, while it is more challenging to adapt to the Sentinel-4 geostationary geometry since there is no OCRA heritage for this type of orbital geometry.