Deep cumulus convection is essential for understanding and modeling of the atmospheric composition, providing a mechanism by which short-lived species can be quickly transported from the planetary boundary layer to the upper troposphere. Because it is a small scale process, highly irregular in time and space, so far most of the scientific community has been skeptical about the legitimacy of zonally averaged representations of tracer convective transport in two-dimensional (2-D) models. To justify and validate such an approach, we derive our parameterization of zonally averaged convective transport from a 3-D model and then validate our 2-D results against the results from that 3-D model using radon 222 and other tracer simulations. We show, both analytically and by comparing 2-D and 3-D model results, that straightforward substitution of zonally averaged parameters into the equations of tracer convective transport produces large errors because the intensity of the convection and the tracer concentration in the updraft and in the environment can be strongly correlated. If such correlations are taken into account in the zonally averaged equations for convective transport, the agreement between 2-D and 3-D results can be significantly improved. We show that our 2-D model with monthly mean convective mass fluxes, properly zonally averaged, is capable of reproducing, with reasonable accuracy, the time series of 222Rn and other tracers obtained from the 3-D simulations. Our convective transport algorithm and input data are available via anonymous ftp. ¿ 1998 American Geophysical Union