Numerical Weather Prediction models require accurate initial conditions and therefore rely heavily on the quality of the assimilation schemes. As known weaknesses currently affect their global analyses of humidity, additional information about the atmospheric water cycle, like rain and cloud observations, is being tentatively introduced in the assimilation schemes.

In this paper, an evaluation of the cloud and rain produced by the ECMWF forecasting system is performed based on Meteosat, ATOVS and SSM/I radiances. Cloud- and rain-retrieved products are compared to the cloud and rain simulated by the forecast model. Also a radiation model is used to generate model-equivalent radiances that are directly compared to the observed ones. It is shown that the cloud and rain structures are well located by the ECMWF model at all latitudes. However, deficiencies are identified and discussed. Finally, prospects for the use of the cloud- and rain- affected radiances in an objective data assimilation system are considered using a one- dimensional variational analysis scheme.