Data from a series of five research flights in marine stratocumulus are used to investigate the structure of radiatively driven, free convective layers. Conditional sampling methods are used to determine the properties of the primary convective elements: negatively buoyant downdraughts. The turbulent velocity data are also subjected to spectral analysis and the turbulent kinetic energy (TKE) balance is evaluated for each case. Results from each case are found to be very similar if scaled using appropriate mixed layer quantities.
Downdraughts are selected using a criterion based on vertical velocity. The distributions of their intersected widths at cloud top, and the velocity spectra, are consistent with downdraughts occupying narrow regions (∼0.1-0.15 h wide, h being the mean mixed layer depth) around the periphery of larger, regular (∼0.5 h-0.75 h diameter) updraughts as suggested by cellular patterns observed in the cloud tops. Downdraughts are found to carry over half of the total turbulent fluxes of heat and water substance within cloud, and also transport variance down the mean gradients. Their thermodynamic properties are consistent with their containing a small fraction of air from above cloud top although their negative buoyancy is almost entirely due to the incorporation of radiatively cooled air. The properties of the downdraughts near cloud top suggest they are formed primarily as a result of the local horizontal convergence of upwelling motions constrained by the inversion. While being cooled radiatively, these quasi-horizontal motions scour the base of the inversion, incorporating some drier air before being forced downwards in relatively narrow zones. Comparisons with other convective layers over land or sea suggest that convection in stratocumulus proceeds with a greater mass flux and correspondingly reduced differences between convective elements and the environment, possibly reflecting the improved ventilation possible at an inversion interface compared with much rougher surfaces.
The TKE balance divides into three main regions and is consistent with the interpretations given above. The pressure transport is implied to be the largest source term close to cloud top. Comparisons with model results reveal some important differences.