Holland, W. R., and A. D. Hirschman, 1972: A numerical calculation of the circulation in the North Atlantic Ocean. Journal of Physical Oceanography, 2, 336-354.

Abstract: A series of numerical experiments are carried out to simulate the three-dimensional circulation in the North Atlantic Ocean and to examine the dynamics therein. The calculations are partly diagnostic in that the density field is not predicted but is given from observations. The main predicted quantities are the velocity and pressure fields.
The results of the basic experiment are compared with observations. The surface currents are quite similar to observations based upon ship draft data, and the surface pressure field is nearly identical to the height of the free surface constructed from a level-of-no-motion hypothesis. The deep pressure variations are nowhere flat or level, however, and the predicted deep currents are quite complex. They are, in fact, strongly controlled by bottom topography and tend to follow f/H contours,where f is the Coriolis parameter and H the depth. The Gulf Stream transport is quite large, reaching a maximum value of 81 x 10⁶ m³ sec^-1, despite the lack of important inertial effects in the western boundary current. Subsidiary experiments show that this large transport value results from an important interaction between the variable density field and bottom topography in the western North Atlantic. When in one experiment the density field was a homogenous one and in another the depth was constant, the maximum transports in the western boundary current were only 14 and 28 x 10⁶ m³ sec^-1, respectively.
Other experiments show that the details of the wind-stress distribution are unimportant when the density field is known; the density field contains most of the information about the long term wind driving. For example, when the wind stress is set equal to zero everywhere (but the density field is maintained in its observed configuration), the Gulf Stream transport is reduced by only 5%. Thus, the pressure torques associated with bottom topography provide the main vorticity input. Finally, it is shown that the results discussed in the basic experiment are not very sensitive to the details of the density field used in the calculation. When these data are highly smoothed and used in a subsidiary calculation, the important feartures, such as enhanced tranpsort in the Gulf Stream and the topographic steering of currents in the deep ocean, are unchanged.