The GFDL general circulation/tracer model has been used to generate the transport coefficients required in two-dimensional (zonally averaged) transport formulations. This was done by assuming a flux-gradient relationship and then, given gradient and flux statistics from two independent (and contrived) model tracer experiments, to derive the coefficients by inversion of this relation. Given the mean meridional circulation from the GCM, the antisymmetric and symmetric parts of the coefficients tensor determine the advective and diffusive contributions to the net meridional transport in the model. The effective transport circulation thus defined differs substantially from the Lagrangian mean and residual circulations and is in fact a simpler representation of the model circulation than either of these. The diffusive component is coherently structured, comprising the following components: (i) Strong quasi-horizontal mixing (Dyy ∼ 1 × 106 m2 s−1) in the midlatitude lower troposphere, apparently associated with fronts and the occlusion of synoptic systems. (ii) A band of strong quasi-horizontal mixing (Dyy ∼ 2 × 106 m2 s−1) stretching across the tropical upper troposphere and the subtropical winter stratosphere. This band follows the band of weak zonal mean winds and is a manifestation in the model of the “surf zone” recently identified by McIntyre and Palmer as a region of breaking planetary waves. Outside the “surf zone,” Dyy ≲ 5 × 105 m2 s−1 in the stratosphere, consistent with other recent estimates. (iii) Intense vertical mixing (Dzz ≳ 10 m2 s−1) in the troposphere at and near the latitudes of the intertropical convergence zone. (iv) Vertical (Dzz ∼ 5−10 m2 s−1) through much of the troposphere, a substantial component of which is associated with subgrid-scale mixing (model convective processes). The validity of the flux-gradient relation as a parameterization of eddy transport processes was tested by implementing the parameterization in a two-dimensional model, using these derived coefficients. In comparison experiments it was found that at the two-dimensional model could reproduce well the zonally-averaged evolution of tracers in the GCM; the quantitative errors that were found may in part result from the finite model resolution, rather than from errors in formulation. Therefore, although the flux-gradient relation is formally justified only in the small-amplitude limit, it appears to be a useful practical description of large-scale transport by finite-amplitude eddies.