Exploiting visible reflectances for estimating cloud parameters and aerosols

Clouds and aerosols remain major sources of uncertainty in numerical weather prediction and climate applications. Reducing these uncertainties requires better observational constraints on key state variables. Visible satellite observations contain rich information on cloud and aerosol properties, yet they are still only marginally exploited in data assimilation systems due to complex and expensive radiative transfer simulations. This study explores the potential of visible reflectances to estimate and constrain cloud and aerosol parameters within the Integrated Forecasting System (IFS) using direct all-sky assimilation in a 4D-Var framework. We demonstrate the assimilation of visible imager observations from various satellite platforms. For clouds, the approach is close to operational readiness. For aerosols, we conducted a proof-of-concept study directly assimilating visible reflectances in cloud-cleared scenes within the IFS-COMPO configuration. We evaluate the impact of visible assimilation on cloud liquid water and ice, as well as on thermodynamic fields. The experiments indicate an improved fit of analyses and short-range forecasts to observed reflectances, several-percent changes in cloud water and ice path, and measurable impacts on temperature and humidity. A case studies of low-level maritime stratus highlight that visible observations can effectively constrain low-level clouds and correct model biases where information from other satellite observations is sparse. The analysis and forecast departures in reflectance space further reveal systematic model biases, offering diagnostic insight into deficiencies in current cloud and aerosol representations. These results represent one of the first demonstrations of direct visible reflectance assimilation in a global 4D-Var system for both clouds and aerosols. At this stage, clouds and aerosols are treated separately, providing an initial step toward broader exploitation of visible observations. Beyond forecasting, the approach offers strong potential for future reanalysis by improving the consistency and realism of long-term cloud records.