The precipitation patterns as simulated with three convection schemes (Betts-Miller, Grell and Kain-Fritsch schemes) are quite different. The convective and stratiform contributions to the total rainfall rates also largely depend on the convection scheme that is used. The 10-day IFA mean rainfall rate however are relatively close. The realism of simulated surface precipitation fields have been further investigated with the help of radar rainfall products, together with an evaluation of the cloud cover as inferred from radiative fluxes. This analysis reveals that the cloud cover is usually underestimated when convective rainfall is significant, whereas it agrees much better with observed for experiments where most of the rainfall occurs on a resolved scale. On the basis of radar data, showing that a large fraction of surface rainfall occurs at a convective scale, this finding suggests a lack of interaction between the convective activity and the cloud cover. Surprisingly, the atmospheric thermodynamical mean states for these various experiments are relatively close to each other, although the contributions from the convective and radiative processes to these mean states show different vertical structures. At the same time, the wind field shows a sensitivity to the convection scheme that is used, through an interplay of complex mechanisms.
Precipitations patterns are not notably modified by changing the radiation scheme (standard radiation scheme, RRTM or CCM2 radiation) or the boundary layer parameterization. Changing these parameterizations however lead to differences of the mean thermodynamical profiles in the same range as those previously found for various convection schemes.