The results of the Phase 1 experiments showed a surprising degree of difference among the participating models, but observations of CO2
arising purely from fossil fuel emissions or seasonal vegetation do not exist. Evaluation of the realism of the various model simulations is therefore impossible. In 1996, the participants agreed to perform additional experiments to “calibrate” the results of TransCom 1, and in addition, seek to understand the mechanisms by which the models diverged so strongly in their results (Denning et al 1999
). With an extremely long atmospheric lifetime, a relatively well-known source, and a twenty year legacy of observations around the planet, SF6 is an ideal trace gas for transport calibration purposes.
While most of the models were reasonably successful at reproducing the "background" observations of SF6, some underestimated marine boundary layer values due to excessive vertical convective transport. Many of the models were less successful at continental locations near sources, where most models significantly overestimated SF6. The “more convective” models matched the observations better at these continental sites than did the “less convective” ones. These results further emphasize the TransCom 1 findings that strong meridional gradients in simulated fossil fuel CO2 at the surface were systematically associated with weak meridional gradients in the upper troposphere, and vice versa.
Although there were distinct differences in the intensity of interhemispheric exchange among the models, these differences could not be adequately understood in terms of spatial distributions of tracer at the surface. Interhemispheric mixing has long been associated with north-south concentration gradients determined from observations, but our results suggested that the meridional gradient measured at the surface is a poor predictor of the true interhemispheric mixing time of a given model.
Both resolved transport and sub-grid scale "column physics" were important in determining the responses obtained by the models. Surprisingly, differences in the subgrid-scale parameterized transport appear to be at least as important in determining model performance as the differences between analyzed winds vs. calculated winds.