An attempt is made to find a steady state of a general circulation model consistent with error estimates of the North Atlantic hydrography during 1981--1985 and with wind and thermodynamic boundary conditions. The quadratic misfit of the model state from the data is minimized using the general circulation model together with its adjoint. A state which is steady within limits of estimated observational error is found, but no such state is simultaneously consistent with the observed hydrographic and surface flux fields. Model dynamics are able to sharpen gradients which were overly smooth in the mapped data, producing a meridional overturning cell with a maximum value of about 21 Sv. At equilibrium, the model must produce its own water masses and has a strong tendency to go toward wintertime conditions. The consequence is an estimated model surface temperature systematically lower than in the hydrography (which comes from all seasons).

The western boundary current thermocline is also both colder and fresher than the mapped hydrography, and the overall meridional heat transport is low (about 0.6¿1015 W). The results suggest that the concept of a realistic steady state North Atlantic circulation has reached the end of its utility, that models with realistic property fluxes and divergences must have much higher resolution, and that the open-ocean boundary conditions must be formulated as control variables. ¿ American Geophysical Union 1993