The effect of convective transport on the global distribution of carbon monoxide (CO) has been examined using NASA's Goddard Earth Observing System Model, version 5 (GEOS5). An ensemble of simulations has been designed to represent the uncertainty introduced by parameterized convection into simulations of atmospheric composition and to isolate the impact of convection from other processes. Rather than perturbing atmospheric state variables, ensemble members were generated by perturbing convective parameters which were identified by a sensitivity analysis as exerting the greatest control over convective transport. GEOS5 simulations were carried out in "replay mode" in which the model is forced by assimilated meteorology, allowing model results to be compared with observed meteorological events. The results of the ensemble of simulations are compared with satellite observations and aircraft observations obtained during NASA's INTEX field campaign to gain insight into optimal parameter settings. Temporal and spatial variability between ensemble members is examined in order to quantify the uncertainty in simulated CO distributions resulting from convective parameterization.