Mesoscale satellite data assimilation: impact of cloud-affected infrared observations on a cloud-free initial model state

This work presents the results of assimilating cloud-affected radiances from geostationary, infrared window and water vapour channels into a mesoscale, cloud-resolving model using a four-dimensional variational assimilation system for the case of an altocumulus cloud over the Great Plains of the United States. In this case, the initial model state, based on reanalysis data, was virtually cloud-free. The impacts of cloudy-scene radiances on a cloud-free model state (and, more generally, accurate satellite observations on inaccurate model initial conditions) in a four-dimensional variational assimilation framework are discussed. Results indicate that, in a cloud-free model state, the assimilation of cloudy radiances modifies the initial conditions as if no cloud exists. This results in a cooling of the surface and lower troposphere upon assimilation of infrared window channels, and an increase in mid-to upper tropospheric humidity upon assimilation of water vapour channels in an attempt to minimize the differences between the modelled and observed radiances. Neither modification of the initial conditions leads to the formation of the observed cloud. The size of the domain and the background error covariance are found to have a significant impact on the results.