Age of air as a diagnostic for transport timescales in global models

Abstract

Abstract. This paper presents the first results of an age-of-air (AoA) inter-comparison of six global transport models. Following a protocol, three global circulation models and three chemistry transport models simulated five tracers with boundary conditions that grow linearly in time. This allows for an evaluation of the AoA and transport times associated with inter-hemispheric transport, vertical mixing in the troposphere, transport to and in the stratosphere, and transport of air masses between land and ocean. Since AoA is not a directly measurable quantity in the atmosphere, simulations of 222Rn and SF6 were also performed. We focus this first analysis on averages over the period 2000–2010, taken from longer simulations covering the period 1988–2014. We find that two models, NIES and TOMCAT, show substantially slower vertical mixing in the troposphere compared to other models (LMDZ, TM5, EMAC, and ACTM). However, while the TOMCAT model, as used here, has slow transport between the hemispheres and between the atmosphere over land and ocean, the NIES model shows efficient horizontal mixing and a smaller latitudinal gradient in SF6 compared to the other models and observations. We find consistent differences between models concerning vertical mixing of the troposphere, expressed as AoA differences and modelled 222Rn gradients between 950 and 500 hPa. All models agree, however, on an interesting asymmetry in inter-hemispheric mixing, with faster transport from the Northern Hemisphere surface to the Southern Hemisphere than vice versa. This is attributed to a rectifier effect caused by a stronger seasonal cycle in boundary layer venting over Northern Hemispheric land masses, and possibly to a related asymmetric position of the intertropical convergence zone. The calculated AoA in the mid–upper stratosphere varies considerably among the models (4–7 years). Finally, we find that the inter-model differences are generally larger than differences in AoA that result from using the same model with a different resolution or convective parameterisation. Taken together, the AoA model inter-comparison provides a useful addition to traditional approaches to evaluate transport timescales. Results highlight that inter-model differences associated with resolved transport (advection, reanalysis data, nudging) and parameterised transport (convection, boundary layer mixing) are still large and require further analysis. For this purpose, all model output and analysis software are available.