A ~ 50 km resolution atmospheric general circulation model (GCM) is used to investigate the impact of radiative interactions on spatial organization of convection, the model’s mean state, and extreme precipitation events in the presence of realistic boundary conditions. Mechanism-denial experiments are performed in which synoptic-scale feedbacks between radiation and dynamics are suppressed by overwriting the model-generated atmospheric radiative cooling rates with its monthly-varying climatological values. When synoptic-scale radiative interactions are disabled, the annual mean circulation and precipitation remain almost unchanged, however tropical convection becomes less aggregated, with an increase in cloud fraction and relative humidity in the free troposphere but a decrease in both variables in the boundary layer. Changes in cloud fraction and relative humidity in the boundary layer exhibit more sensitivity to the presence of radiative interactions than variations in the degree of aggregation. The less aggregated state is associated with a decrease in the frequency of extreme precipitation events, coincident with a decrease in the dynamical contribution to the magnitude of extreme precipitation. At regional scales, the spatial contrast in radiative cooling between dry and moist regions diminishes when radiative interactions are suppressed, reducing the upgradient transport of energy, degree of aggregation and frequency of extreme precipitation events. However, the mean width of the tropical rain belt remains almost unaffected when radiative interactions are disabled. These results offer insights into how radiation-circulation coupling affects the spatial organization of convection, distributions of clouds and humidity, and weather extremes.

East Africa has two rainy seasons: the long rains (March–May, MAM) and the short rains (October–December, OND). Most CMIP3/5 coupled models overestimate the short rains while underestimating the long rains. In this study, the East African rainfall bias is investigated by comparing the coupled historical simulations from CMIP5 to the corresponding SST-forced AMIP simulations. Much of the investigation is focused on the MRI-CGCM3 model, which successfully reproduces the observed rainfall annual cycle in East Africa in the AMIP experiment but its coupled historical simulation has a similar but stronger bias as the coupled multimodel mean. The historical−AMIP monthly climatology rainfall bias in East Africa can be explained by the bias in the convective instability (CI), which is dominated by the near surface moisture static energy (MSE) and ultimately by the MSE’s moisture component. The near surface MSE bias is modulated by the sea surface temperature (SST) over the western Indian Ocean. The warm SST bias in OND can be explained by both insufficient ocean dynamical cooling and latent flux, while the insufficient short wave radiation and excess latent heat flux mainly contribute to the cool SST bias in MAM.