We perform simulations using global ocean general circulation models of the uptake by the ocean and the transport within the ocean of 14C from atmospheric nuclear bomb tests. The models used are (1) a modified version of the Geophysical Fluid Dynamics Laboratory modular ocean model (MOM) configured with simple horizontal subgrid-scale mixing of tracers; (2) the same model using the Gent and McWilliams (1990) subgrid-scale tracer transport parameterization, which includes isopycnal mixing; and (3) the ocean isopycnal (OPYC) model of J. M. Oberhuber, which uses isopycnal subgrid-scale mixing of tracers within an isopycnal coordinate system. In addition, we perform another MOM simulation which illustrates the sensitivity of the results to the thickness of the surface mixed layer. To minimize differences in the results due to differences in forcings, all our simulations use the same formulations for surface fluxes of heat and fresh water and the same sea ice model. In addition, the numerical grids are all nearly the same. Comparison of our model results to revised estimates of oceanic bomb 14C from the Geochemical Ocean Sections Study observations and to annual mean temperature observations shows that our results are sensitive to the models' treatments of subgrid-scale mixing of tracers and to the thickness of the surface mixed layer.

Since the simulated vertical profiles of bomb 14C or temperature alone can be brought close to the observed profile by adjusting model parameters, we judge the realism of modeled vertical transport by how well vertical profiles of both bomb 14C and temperature can be simulated with one set of parameter values. By this standard, OPYC does better than our baseline configuration of MOM, which we attribute to the unrealistic treatment of subgrid-scale mixing of tracers and of the surface mixed layer in this simulation. We illustrate this by showing that the use of the Gent and McWilliams tracer transport parameterization results in a markedly improved vertical temperature profile and almost no change in the bomb 14C distribution, compared with our baseline MOM simulation. Finally, we show that increasing the effective mixed layer thickness in MOM from 30 to 76 m results in a slightly more realistic vertical profile of bomb 14C but a slightly less realistic temperature profile. ¿ American Geophysical Union 1995