This thesis describes two inverse models solving for a quasi-stationary ocean circulation, and discusses the circulation in the North Atlantic as derived from them. They include finite-element inverse section model FEMSECT (Losch et al., 2004) and 3D inverse finite-element ocean model IFEOM. Both models are based on the adjoint technique and use finite-element discretization to accurately represent the sloping bottom topography. FEMSECT exploits the thermal wind relation, and seeks for a compromise in the least square sense between the hydrographic and mooring data. Its control parameters are the reference velocities and hydrographic fields. Its novel feature is the ability to take into account the bottom triangles. The inverse finite element ocean model respects the continuity locally and globally and also exploits the flexibility of 3D finite element grids. It is based on a steady-state version of the finite element ocean general circulation model FEOM (Danilov et al., 2004a). The IFEOM solves for density by minimizing the misfit between it and the density data under strong momentum and weak potential density balance constraint. An additional deep pressure gradient constraint (below 2000 m) is suggested and shown to be crucial for keeping the integral properties of the diagnosed ocean circulation close to those of the forward run of FEOM. The circulation in the North Atlantic is estimated by assimilating several data sets. The results are encouraging and indicate that IFEOM can be used to assimilate a climatological circulation from high quality hydrographic measurements.Keywords: data assimilation, adjoint method, North Atlantic circulation, finite elements