Understanding the mechanisms that control the diurnal phase of tropical convection using a hierarchy of models.

Although global climate and forecast models have improved their representation of the diurnal phase of convection, biases remain—especially in the tropics and over land. For example, many models experience precipitation peaks in phase with surface buoyancy fluxes, instead of several hours later as observed. The extent of the consequences of these convection phase errors is not well understood, as the processes involved interact in a nonlinear way. Moreover, convection is crucial for maintaining the thermal and moisture profile of the tropical atmosphere; therefore, it is important not only to further improve the representation of convection in models, but also to fundamentally understand the mechanisms that exert control on the diurnal phase of convection.

Controlled simulations over a wide parameter space with models of varying complexity provide a unique opportunity to further understanding of this problem. More specifically, here we use observations, reanalysis, a single-column model, a 2D model, and a new LES model in an interdependent way. Simulations are initialized and forced with diurnal radiative forcing, which closely matches observations, and are run into radiative-convective equilibrium (RCE). In these RCE simulations, we analyze the extent to which the diurnal phase can be explained in the absence of mesoscale circulations.

Simulations that separately vary surface thermal inertia, evaporation rates, and longwave/shortwave clear-sky and cloud feedback are conducted. These simulations show that although a simple radiation scheme coupled to a surface with the correct thermal inertia may be sufficient to distinguish between tropical maritime and continental convection, nonlinear cloud-radiative feedback may be important for introducing the lag between the peaks in precipitation relative to surface buoyancy fluxes over homogeneous continental surfaces. Previously, surface heterogeneity and mesoscale convective organization have been linked to phase lags of deep convection. We show that mesoscale organization is not necessary to obtain the observed phase lags and that cloud feedbacks may play a role in delaying convection over a homogeneous continental surface.