This article presents the main features of a three-dimensional model for deep convection developed with special care given to the formulation of subgrid turbulent processes. It explicitly simulates the dynamics of turbulent eddies, including condensation and precipitation processes. Second-order moments are expressed as a function of the grid-averaged field of variables and of a prognostic turbulent kinetic energy. The formulation includes a simple statistical treatment of subgrid condensation and subgrid conversion of cloud water into rain water. The coherence and relative importance of the various closure hypotheses are tested in an idealized case of precipitating cloud. Results indicate the extent that features of the computed field are dependent on hypotheses used in the turbulence closure, choice of the basic turbulent variables, and formulation of the second-order moments. Significant benefits are obtained from the use of variables that are conserved in the condensation process. The computa... Abstract This article presents the main features of a three-dimensional model for deep convection developed with special care given to the formulation of subgrid turbulent processes. It explicitly simulates the dynamics of turbulent eddies, including condensation and precipitation processes. Second-order moments are expressed as a function of the grid-averaged field of variables and of a prognostic turbulent kinetic energy. The formulation includes a simple statistical treatment of subgrid condensation and subgrid conversion of cloud water into rain water. The coherence and relative importance of the various closure hypotheses are tested in an idealized case of precipitating cloud. Results indicate the extent that features of the computed field are dependent on hypotheses used in the turbulence closure, choice of the basic turbulent variables, and formulation of the second-order moments. Significant benefits are obtained from the use of variables that are conserved in the condensation process. The computa...