Abstract By forcing a third-generation wave-prediction model with surface stresses from the European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric model, it was discovered that lower wave heights were generated than by forcing with the ECMWF surface winds. The apparent inconsistency between surface stresses and surface winds in the atmospheric model turns out to be time-step dependent. A similar conclusion may be inferred from results of the WAMDI group. Apparently, a number of atmospheric models have inaccuracies in the boundary-layer scheme near the surface. In this paper it is argued that the reason for the inaccuracies is related to the numerical integration scheme that is used in these models. It is shown that a numerical scheme that treats physics and dynamics separately has an equilibrium that is time-step dependent. An alternative scheme—namely, simultaneous, implicit treatment of both physics and dynamics—removes this deficiency. Possible consequences for atmospheric-, wave-, and ...