Report on improved retrievals for IFS-AER and evaluation of global CAMS products. D1.7

Aerosols play a crucial role in many atmospheric processes such as chemistry, radiation balances, and cloud formation. For air quality, aerosols are important because of their adverse effects on human health. The wavelength-dependent interactions between aerosols and radiation are quantified by column-integrated bulk optical properties like aerosol optical depth (AOD), single scattering albedo (SSA), and asymmetry parameter (ASY), as well as three dimensional diagnostics such as extinction and backscatter coefficients. The ECMWF's Integrated Forecasting System (IFS) provides these diagnostics in forecasting mode and assimilates satellite-retrieved AODs at 550 nm from MODIS and VIIRS (since 2023) as well as from AATSR Envisat (2003-2012) in a reanalysis system. Based on modelled aerosol mass mixing ratios (MMRs), the optical diagnostics are computed. Incorporating observations aims to bring modelled results closer to reality. For optical diagnostics, it is possible to align measured and modelled results for assimilated quantities, such as AOD, while introducing greater discrepancies in underlying fields, such as aerosol concentrations. For example, when modelled aerosol fields are optically less active than in reality, total aerosol concentrations or emissions are perturbed to match ground and satellite retrieved AODs. The existing assimilation schemes generally rely on limited input data such as single-wavelength AOD, which does not contain sufficient information on aerosol speciation. In addition, various parameterizations for individual aerosols like size, scattering properties and mixing state contributes to model uncertainties, presenting difficulties in translating optical observations to the model state of mass concentration. Therefore, it is crucial to have an accurate optical representation of aerosols to maximize the benefit of assimilating optical diagnostics, evaluate sensitivities of various aerosol characteristics, and quantify the model uncertainties in time and space.