Towards a Fast, Physically-Based Solar Radiance Model for Cloud and Aerosol Retrievals and Assimilation

Solar radiances are amongst the most useful satellite observations available for characterizing clouds and aerosols, but their use in data assimilation and synergistic retrievals is hampered by the lack of a suitably fast radiance model that can cope with an arbitrary profile of different particle types. In this presentation we describe the "Forward Lobe Two-Stream Radiance Model" (FLOTSAM), which has the potential to fill this gap. It exploits the fact that particle scattering patterns are frequently dominated by a "forward lobe" of width around 30 degrees, which enables the radiation field at a given height to be described by the sum of the direct (unscattered) solar beam, a wide beam proagating away from the sun due to scattering by the forward lobe, and a diffuse radiation field described by the two-stream equations. Differential equations are formulated and solved for these components of the radiation field, and the resulting flux profiles for each component are validated using Monte Carlo calculations in which photons are labeled according to which component they belong to. This description of the radiation field is then used to compute the source function for a radiance calculation towards the satellite. Evaluation against DISORT and Monte Carlo simulations show that currently FLOTSAM is able to compute radiances with an RMS error of around 3% for aerosol profiles and somewhat larger for cloudy profiles. FLOTSAM is also easily fast enough to use in data assimilation and retrievals: a 40-layer description of the atmosphere can be computed in around 30 microseconds, and the corresponding Jacobian in 120 microseconds. The figure shows a comparison of a geostationary VIS satellite image and the corresponding FLOTSAM calculation on the ECMWF model cloud fields at the same time. Can you tell which is which?