This paper examines how satellite altimeter and scatterometer measurements could be jointly used in a numerical ocean model in an attempt to simulate a realistic ocean circulation. The aim of the study is to determine quantitatively how sensitive the efficiency of the assimilation process may be to the variability of the wind stress curl and to evaluate the ability of the Topex/Poseidon altimetric data to correct for the sampling of scatterometer wind data that is to be provided by the forthcoming ERS 1 and NSCAT satellite missions. The model is quasigeostrophic, eddy-resolving and multi-layered, and is applied to the prognostic description of the mid-latitude ocean circulation in a schematic box ocean. Satellite altimeter data are simulated from model runs (forced by ECMWF winds) under the conditions of the Topex/Poseidon mission and are assimilated into the model by using a Newtonian relaxation or nudging technique. Satellite scatterometer wind data are simulated from ECMWF daily analyses of the wind stress over the North Atlantic. The use of realistic winds varying over a large range of temporal and spatial scales clearly has important dynamical consequences for the resulting ocean circulation but also militates against the efficiency of the altimeter data assimilation. The insertion of the altimeter data into the model is still able to constrain strongly the slow baroclinic circulation. However, the high frequency forcing generates barotropic Rossby waves that dominate the instantaneous flow throughout the basin and prevent complete convergence in the barotropic mode. Nevertheless, by smoothing out this fast time variability it is demonstrated that the assimilation of the altimeter data is still efficient in driving the larger time scales of the model ocean circulation. When performing altimeter data assimilation into a model domain forced by scatterometer wind data, the efficiency of the assimilation process is only slightly diminished but remains globally equivalent to the true wind forcing situation. Different ways of assimilating altimeter data were tested; they revealed that the use of the sea-level instead of the vorticity is more favorable; this result was not observed when model forcing did not incorporate such variability, i. e. when it was schematized by large scale constant wind stress curl (Verron, 199 1).