Data from multiple satellite and in situ sources are used to investigate the dominant raining cloud populations in the tropics, with the purpose of documenting how diverse the raining cloud populations are at any given time over a scale similar in size to the grid-box (∼100 – 200 km) of a present-day global climate model (GCM). For all locations in the tropics, three similar rainfall clusters (defined according to their ensemble of clouds) are found. Differences in mean-state rainfall (e.g. East versus West Pacific Ocean) are largely the result of similar rainfall clusters occurring at ocean basin-dependent relative frequencies of occurrence.

Area-average rainfall rates are substantially different for each cluster. While each rainfall cluster is observed in all tropical basins, differing relative frequencies of occurrence imply that rainfall lifecycles (i.e. the time duration for transition from light to deep rainfall) vary as a function of basin. Among the processes influencing this transition, both mesoscale cold pools (inferred from QuikSCAT surface wind field retrievals) and convective inhibition (CIN, derived from radiosonde-observations) emerge as important parameters driving the transition from light rainfall to deep convection at the spatial scale of 100 – 200 km. Associated with significant increases in rainfall are substantial decreases (40%) in convective available potential energy (CAPE).

The temporal evolution of rainfall clusters is derived for different lifecycle stages of a composite Madden-Julian Oscillation (MJO) event. It is found that the rainfall cluster consisting of shallow (<3 km) and congestus raining clouds exhibits little temporal variation for all stages of the composite event, while non-raining scenes and deeper clouds are modulated as a function of time for all stages. Instead of a “transition” from shallow to deep convection, the results suggest an “addition” of deep convection at the expense of non-raining scenes. Unique to the initiation stage, deep organized convective systems are rare until 1 – 5 days before the development of a convective anomaly that finally begins propagating eastward. The lack of deep convection during the initiation stage relative to other stages is associated with both decreased values of columnar water vapor (TPW) and increased stability in the lower-troposphere. Both are hypothesized to preclude the development of deeper convection, thus allowing for the slow (10-30 day) increase in TPW by evaporation to continue, in contrast to later stages of the MJO when moisture convergence serves as the largest contributor to moistening.

The analyses described above are applied to output from a novel multiscale-modeling framework (MMF) coupled with a slab ocean model. The extent to which the MMF yields results similar to the observational depictions outlined above is discussed in great detail.