Clouds are critical regulators of the earth's weather and climate. They strongly affect the earth's radiation budget, control latent heat release and soil water availability, and even are important sites where chemistry takes place. A growing number of recent studies emphasize the interest of explicitly representing cloud-water phases and microphysics in atmospheric models, both for climate modeling and weather forecasting. Yet, because of their complexity, and the vast range of space and time scales over which they operate, it is difficult to validate cloud predictions by atmospheric models, and even more so to assess the cloud parameterization assumptions.

In this paper, an evaluation of the explicit cloud scheme of Meso-NH, a mesoscale model, is presented. We adopt the ``model-to-satellite'' approach, in which satellite brightness temperature (BT) images are directly compared to synthetic BTs computed from predicted model fields. First, this approach offers the advantage that the satellite BTs are used in an independent and objective way, without any combination and treatment using ancillary data or assumption. Second, the observations from satellites are easily available and offer a good coverage in time and space. Last but not least, the observations integrate the spatial variability of the microphysical processes at a scale directly comparable with the gridmesh of a mesoscale model. Thus, no further hypothesis or data reduction for scale extension is needed.

Three different meteorological situations are examined, showing that the model is able to simulate realistic synthetic BTs, both in mid-latitude and in subtropics, with horizontal resolutions ranging from 75 km to 12 km. Moreover, the model-to-satellite approach, which combines an explicit cloud scheme implemented in a mesoscale model with a detailed radiative transfer code, gives access to the tuning of the ice parameterization. A comparison made with three different values of the ice-to-snow autoconversion threshold shows a significant improvement of the synthetic BTs, and suggests an optimum value minimizing the difference between simulated and observed BTs. A similar test conducted on the ice water path and the liquid water path shows a good agreement with retrievals from the SSM/T2 and SSM/I observations, whatever the autoconversion threshold chosen.