On the estimation of transport characteristics of atmospheric data sets

Transport characteristics of an atmospheric general circulation model have been estimated byanalysing 30 000 trajectories over several months. These trajectories have been used to constructan advection-diffusion operator for the zonally-averaged transport of passive tracers. Severaldifferent methods have been tried for this purpose. The most promising seems to be a methodwhere eddy-flux statistics for a large number of tracers are generated and used to find anoptimized flux-gradient relationship. An alternative approach based on Lagrangian averagingis shown to be much less robust with respect to changes in the details of the procedure. Themeridional circulation of the advection-diffusion operator is direct almost everywhere, but isvery weak outside the tropics. Diffusion coefficients show a quite complicated structure withminima near the jet maxima. Certain limitations of the advection-diffusion model have beenidentified. In particular, there seems to be problems related to Walker cells. The use of a largenumber of different tracers has also allowed us to improve on the flux-gradient relationship byincluding higher order terms. The relevance of the advection-diffusion approach for the transportof potential vorticity has been tested. The Stokes drift is shown to be very important forpotential vorticity fluxes near the tropopause.