An analysis of the impact of convective parameter sensitivity on simulated global atmospheric CO distributions

In an effort to better understand how uncertainty in simulated convection propagates into simulations of global trace gas distributions, we have constructed an eight‐member ensemble of simulations using NASA's Goddard Earth Observing System Version 5 (GEOS‐5) general circulation model (GCM). The ensemble was created by perturbing parameters in the model's moist physics schemes found to strongly influence the magnitude of convective mass flux. Globally, ensemble spreads in column CO are typically small (less than 4% of the mean column value) and, in many areas, are not significantly different from internal model variability. The largest ensemble spreads are found near source regions and outflow pathways. At the majority of remote surface monitoring sites, the annual mean ensemble spread is less than 5%, indicating that these locations, which are often the basis of inversion studies, are relatively insensitive to uncertainty in the representation of convection. We also examine in greater detail two simulations in which the magnitude of convective mass flux is significantly altered. Changes to convective parameters strongly influence grid‐scale vertical and turbulent transport processes in addition to convective mass flux. Despite large differences in the magnitude of convective mass fluxes, this compensating behavior by other model processes results in comparable atmospheric residence times in the two simulations and largely similar global CO distributions. The results indicate that convective mass flux is strongly related to other vertical transport processes in a GCM and cannot be viewed as entirely separate. Future studies of the role of convective transport need to consider the relationship between convective and total mass flux.