A global spectral model extending approximately from 15 up to 120 km is used in both diagnostics and simulations of the observed monthly middle atmospheric climatology as represented in the CIRA-86 empirical model. First, the model in a two-dimensional mode is initialized with the empirical temperature and wind distributions for each month. An analog of the nudging data assimilation technique is then applied to retrieve zonal mean zonal momentum depositon rates necessary to keep the model's state in the vicinity of the observed climatology. This is combined with a straightforward procedure for estimating the global mean vertical diffusion coefficient from the thermodynamic equation averaged globally for each month. These estimates agree surprisingly well with recent observations in the mesosphere and lower thermosphere. At the second stage the inferred eddy diffusivities and zonal accelerations that are commonly believed to be produced by dissipating and breaking atmospheric waves of different scales are used as input in annual cycle integrations of the model. For the first time quantitative comparisons of the simulated and initial diagnosed climatologies are presented. The simulation results obtained using the diagnosed zonal forcing and vertical diffusion coefficients compare favorably with the empirical model on the one hand and with simulations using a gravity wave parameterization on the other. In particular, the empirical temperature model is reproduced with an annual global rms temperature deviation of 3.2 K or 2% in the 15- to 110-km layer. ¿ American Geophysical Union 1994