A modification of the classical Ekman model of oceanic wind-driven currents including the Stokes drift and stratification effects is discussed. The modification is formulated as an application of turbulence mechanics accounting for the curvature effect of velocity fluctuation streamlines. It is shown that similar to the Stokes drift effect, the presence of a density jump layer (pycnocline) decreases the veering of the flow velocity vector at the surface from the direction of the wind stress. It is shown also that in the pycnocline the decrease of the norm of the velocity vector as well as its rotation with depth is smaller than in the regions adjacent to the pycnocline. If the Stokes drift and stratification effects are neglected, the model reduces to the classical Ekman solution with the coefficient of the turbulent shear viscosity replaced by an effective viscosity coefficient. The vertical distributions of velocity predicted by the modified model are compared with the velocity data measured in the Drake Passage and within the Long-Term Upper Ocean Study (LOTUS) in the North Atlantic.