Numerical experiments with a two-dimensional cloud-resolving model are reported in this paper. Atmospheric data from the Tropical-Ocean Global Atmosphere Coupled Ocean-Atmosphere Response Experiment (TOGA-COARE) were used to provide realistic atmospheric vertical profiles and advective forcing. Two types of boundary conditions were imposed to the cloud-resolving model, in order to investigate the effects of inhomogeneous sea surface temperatures (SST) in the statistical behavior of clouds and convective systems. First, idealized sinusoidal perturbations with a 1K amplitude and various wavelengths were added to the observed SST field in a set of simulations. Most of the collective, model-domain averaged characteristics of convection were almost insensitive to the SST changes. However, Fourier transforms of the modeled fields showed that certain scales of convective activity can be favored depending on the spatial scale of the surface forcing, with a maximum of convective rainfall occurring for sinusoidal SST perturbations of ~ 100km wavelengths. Such a spatial scale is possibly associated with the distance of two convective cores with a subsidence zone in between. Second, the cloud-resolving model was coupled to a dynamic ocean model in order to attain realistic two-way feedbacks. Attention was given to the simulated inhomogeneous SST field in the ocean model, associated with precipitation-produced freshwater lenses and cloud shadows and how they influence the atmospheric variables. It was found that significant contributions for the SST come from typical freshwater lens scales (despite an amplitude smaller than 1K, as used in the uncoupled simulations) and from scales corresponding to typical distances between one freshwater lens and another. The results suggest a weak positive feedback between the inhomogeneous sea surface and convective clouds favoring certain scales of convective organization (namely the ones of the order of ~100km).