This talk presents three-dimensional numerical simulations of oceanic trade cumulus clouds underlying stratocumulus clouds. They are based on a case studied in a Global Energy and Water Experiment Cloud System Study (GCSS) model intercomparison that is loosely based on observed conditions during the Atlantic Trade Cumulus Experiment (ATEX). It is motivated by the importance of this cloud type to global cloud radiative forcing, and their role as a feeder system for deep convection in the tropics. This study focuses on the sensitivity of the modeled cloud field to the domain size and the grid spacing. Domain widths from 6.5 to 20 kilometers and horizontal grid spacings ranging from 10 to 80 meters, with corresponding vertical grid spacing ranging from 5 to 40 m, are studied, involving massively parallel computations on up to 2.5 billion grid-cells. The combination of large domain size and small grid resolution provides an unprecedented perspective on this type of convection.

The mean stratocumulus cloud fraction, optical depth, and vertical fluxes of heat, moisture and momentum are found to be quite sensitive to both the domain size and the resolution. The sensitivities are associated with a strong feedback between cloud fraction, cloud-top radiative cooling, and entrainment. The properties of individual cumulus clouds rising into the stratocumulus are less affected than the stratocumulus clouds. The simulations with 80 m horizontal/40 m vertical resolution are clearly under-resolved, with distinctly different distributions of liquid water within the clouds. Increasing the resolution to finer than 40 m horizontal/20 m vertical affects the inversion structure and entrainment processes somewhat, but has less impact on the structure of individual clouds. Large-domain simulations exhibit mesoscale structure in the cloud organization and liquid water path. This mesoscale variability feeds back on the domain-mean properties through the cloud-radiative feedback. These simulations suggest that very large computations are required to obtain meaningful cloud statistics for this case.