Search

Climate models consistently predict an acceleration of the Brewer-Dobson circulation (BDC) due to climate change in the 21st century. However, the strength of this acceleration varies considerably among individual models, which constitutes a notable source of uncertainty for future climate projections. To shed more light upon the magnitude of this uncertainty and on its causes, we analyse the stratospheric mean age of air (AoA) of 10 climate projection simulations from the Chemistry-Climate Model Initiative phase 1 (CCMI-I), covering the period between 1960 and 2100. In agreement with previous multi-model studies, we find a large model spread in the magnitude of the AoA trend over the simulation period. Differences between future and past AoA are found to be predominantly due to differences in mixing (reduced aging by mixing and recirculation) rather than differences in residual mean transport. We furthermore analyse the mixing efficiency, a measure of the relative strength of mixing for given residual mean transport, which was previously hypothesised to be a model constant. Here, the mixing efficiency is found to vary not only across models, but also over time in all models. Changes in mixing efficiency are shown to be closely related to changes in AoA and quantified to roughly contribute 10 % to the long-term AoA decrease over the 21st century. Additionally, mixing efficiency variations are shown to considerably enhance model spread in AoA changes. To understand these mixing efficiency variations, we also present a consistent dynamical framework based on diffusive closure, which highlights the role of basic state potential vorticity gradients in controlling mixing efficiency and therefore aging by mixing.

The stratospheric age of air (AoA) is a useful measure of the overall capabilities of a general circulation model (GCM) to simulate stratospheric transport. Previous studies have reported a large spread in the simulation of AoA by GCMs and coupled chemistry–climate models (CCMs). Compared to observational estimates, simulated AoA is mostly too low. Here we attempt to untangle the processes that lead to the AoA differences between the models and between models and observations. AoA is influenced by both mean transport by the residual circulation and two-way mixing; we quantify the effects of these processes using data from the CCM inter-comparison projects CCMVal-2 (Chemistry–Climate Model Validation Activity 2) and CCMI-1 (Chemistry–Climate Model Initiative, phase 1). Transport along the residual circulation is measured by the residual circulation transit time (RCTT). We interpret the difference between AoA and RCTT as additional aging by mixing. Aging by mixing thus includes mixing on both the resolved and subgrid scale. We find that the spread in AoA between the models is primarily caused by differences in the effects of mixing and only to some extent by differences in residual circulation strength. These effects are quantified by the mixing efficiency, a measure of the relative increase in AoA by mixing. The mixing efficiency varies strongly between the models from 0.24 to 1.02. We show that the mixing efficiency is not only controlled by horizontal mixing, but by vertical mixing and vertical diffusion as well. Possible causes for the differences in the models' mixing efficiencies are discussed. Differences in subgrid-scale mixing (including differences in advection schemes and model resolutions) likely contribute to the differences in mixing efficiency. However, differences in the relative contribution of resolved versus parameterized wave forcing do not appear to be related to differences in mixing efficiency or AoA.