TROPOMI’s Aerosol Layer Height Algorithm: Interpretation of the retrieved height parameter

TROPOMI’s Aerosol Layer Height algorithm currently assumes a simplified aerosol profile in the form of a single layer of homogeneously distributed aerosols. The height of this layer is retrieved by using height information from the O2 A band. It is not straightforward to interpret the retrieved height parameter, since true aerosol profiles are typically complex and heterogeneous. We used radiative transfer theory to better understand the retrieved parameter. In addition to an available multiple scattering description of the radiation field, a single scattering approximation was developed as well. Comparing the single scattering approximation to the full multiple scattering radiative transfer revealed that the approximation is accurate inside strong absorption parts of the O2 A band for realistic aerosol profiles. Retrieval simulations for realistic aerosol scenes suggested that the retrieved height parameter can be interpreted as a centroid of the aerosol profile, if the surface albedo is not fitted. The retrieved height parameter was compared to three different centroids of the aerosol extinction profile, which are constructed using weighted averages. Two of these methods take into account the sensitivity of the measured top of atmosphere radiance with respect to aerosol at different altitudes. This sensitivity was obtained by taking a directional derivative of a functional, which maps the aerosol extinction profile to a measured top of atmosphere radiance. This yields semi-analytic derivatives. The two methods differ in the way the derivatives are determined. The first uses a multiple scattering description and the other uses the single scattering approximation. All centroid methods showed good agreement with the retrieved height parameter. In additional retrieval experiments we also fitted the surface albedo. These experiments showed that, as soon as the standard error of the surface albedo exceeds 10−3, a bifurcation occurs, which results in unrealistic results. These unrealistic results can no longer be interpreted as centroids of aerosol extinction profiles.