Impact of Shallow Cumulus on Cloud-Permitting WRF Simulations of the Diurnal Cycle of Deep Convection over the Amazon

Understanding and representing the diurnal cycle of convection in climate models remains one of the most significant challenges in climate science. In this study we examine the radiative impact of shallow cumulus clouds on a month-long cloud-permitting (Dx=1 km) WRF simulation of convection over the Amazon region observed during the GoAmazon2014/5 field campaign. Compared to observations collected by the DOE Atmospheric Radiation Measurement Mobile Facility (AMF), the control simulation largely underestimates the frequency of shallow cumulus clouds during daytime and overestimates the frequency of propagating deep convection during nighttime/morning. These biases result in erroneous cloud radiative effects and a positive bias in downwelling shortwave radiative flux at the surface. This overestimation of shortwave heating in turn drives excessive surface latent and sensible heat fluxes in the model. As a result, the simulated deep convection is too intense, precipitates too much in the morning with an incorrect diurnal cycle compared to the observations. We use the AMF observations to develop a simplified shallow cumulus cloud parameterization that mitigates the bias in shallow cumulus shortwave cloud radiative effects. We examine the how the improved surface energy budget impacts the simulated diurnal cycle of deep convection. Our results highlight the need for shallow cumulus parameterizations at sub-10km resolution for future climate models, and the importance of cloud-radiative interactions and land-atmosphere feedbacks on the diurnal cycle of convection in this region.