A statistical cloud parameterization is presented, where in addition to earlier work the subgrid moisture/thermal variability is diagnostically determined as a sum of a turbulent and a convective contribution, the latter being a function of the convective mass flux. It is shown that the inclusion of a convective contribution allows us to reasonably reconstruct the time mean and variability of the variance of the conserved quantities total water and liquid water static energy. Application of the parameterization in a regional model during a 30-day period of the Tropical Convection, Cirrus and Nitrogen Oxides (TROCCINOX) experiment over southern Brazil demonstrates a clear improvement in the representation of the cloud fields. The subgrid cloud parameterization significantly reduces both the biases and phase between the observed and simulated brightness temperatures in the thermal infrared band and the observed and simulated precipitation intensities, resulting in an improved representation of the diurnal cycle of convection over land. The strong impact on the diurnal cycle is mainly a consequence of reduced surface fluxes due to cloud shading and a change in tropospheric stability due to additional condensation/sublimation.