A first step for improving the climatological state of high-resolution general circulation models by means of data assimilation is presented. A method developed for the assimilation of statistical characteristics into chaotic ocean models is applied to assimilate SSH variability from TOPEX/POSEIDON and ERS1 in association with temperature and salinity from the World Ocean Atlas 1997 in order to estimate the underlying mean circulation. The method requires a parameterization of SSH variability which derives from the approach of Green and Stone. By estimating initial conditions for temperature and salinity, a mean state is achieved which, although not fully consistent with the altimetric and climatological data, is markedly improved on time scales of one year in comparison to the control run. The assimilation of SSH variability data introduces complementary information about the main frontal structures consistent with climatological observations. The state is however not an equilibrium state and returns back to the first guess quasi-equilibrium state for longer integration periods.