Impact of 3D cloud structures on tropospheric NO2 column measurements from UV-VIS sounders

Operational retrievals of tropospheric trace gases from space-borne spectrometers are made using 1D radiative transfer models. To minimize cloud effects generally only partially cloudy pixels are analysed using simplified cloud contamination treatments based on radiometric cloud fraction estimates and photon path length corrections based on oxygen collision pair (O₂-O₂) or O₂A-absorption band measurements. In reality, however, the impact of clouds can be much more complex, involving scattering of clouds in neighbouring pixels and cloud shadow effects. Therefore, to go one step further, other correction methods may be envisaged that use sub-pixel cloud information from co-located imagers. Such methods require an understanding of the impact of clouds on the real 3D radiative transfer. We quantify this impact using the MYSTIC 3D radiative transfer model. The generation of realistic 3D input cloud fields, needed by MYSTIC (or any other 3D radiative transfer model), is non-trivial. We use cloud data generated by the ICOsahedral Non-hydrostatic (ICON) atmosphere model for a region including Germany, the Netherlands and parts of other surrounding countries. The model simulates realistic liquid and ice clouds with a horizontal spatial resolution of 156 m and it has been validated against ground-based and satellite-based observational data.

As a trace gas example, we study NO₂, a key tropospheric trace gas measured by the atmospheric Sentinels. The MYSTIC 3D model simulates visible spectra, which are ingested in standard DOAS retrieval algorithms to retrieve the NO₂ column amount. Spectra are simulated for a number of realistic cloud scenarios, snow free surface albedos, and solar and satellite geometries typical of low-earth and geostationary orbits. The retrieved NO₂ vertical column densities (VCD) are compared with the true values to identify conditions where 3D cloud effects lead to significant biases on the NO₂ VCDs. A variety of possible mitigation strategies for such pixels are then explored.