The capability of 4D-Var systems to assimilate cloud-affected satellite infrared radiances

Four-dimensional variational (4D-Var) assimilation schemes assume the linearity of their forward model ill the vicinity of prior information and usually do not properly handle variables that have finer temporal and spatial scales in the real world than in the forward model. Hence cloud-affected satellite infrared radiances are discarded from numerical weather-prediction 4D-Var systems. despite the critical need of observations within the cloudy regions. This paper suggests the reappraisal of that choice, subject to achieving improvements in the numerical simulation of cloudiness. A new observation operator. that computes cloud-affected infrared radiances from 4D-Var control variables. namely atmospheric temperature, humidity, ozone, surface temperature and surface pressure, is presented. The vertical distributions of cloud cover and of cloud condensate are diagnosed in the operator itself. The goal of this paper is to assess the feasibility of using it to assimilate cloud-affected infrared radiances. such as those from the narrow-band Advanced Infrared Sounder on-board the Aqua platform or those from the broad-band Meteosat Visible and Infrared Imager. It is shown that there is a potential benefit in assimilating directly in 4D-Var some of the upper-tropospheric channels at 4.5, 6.3 and 14.3 mum in the presence of clouds, for instance the 6.3 mum channel oil board all the geostationary satellites. The approach is illustrated with one-dimensional variational retrievals collocated with radiosonde observations.