A prognostic model based on assimilation of meteorological observations is proposed for the air quality control in an urban area. A short-term forecast of the urban weather is obtained using a one-dimensional unsteady model of the atmospheric boundary layer. High reliability of meteorological forecast is achieved due to assimilation of meteorological data (wind velocity, temperature, and humidity) in the model. The model predicts a detailed pattern of the vertical distribution of horizontal wind components and turbulent parameters, which are very important in simulating the pollution dispersion in the near-surface air. Pollution transport through an urban air basin is predicted based on the equations of spatial transport of harmful species chosen for study. The following pollution sources are considered: stacks of industrial and heat-and-power production plants (point sources), traffic (linear sources), and large industrial area (area sources). The problem is solved numerically using the finite volume method and the factorization method. To speed up the solution of the problem, we used high performance computers of TSU and IAO SB RAS. The parallel computational algorithm used for numerical solution is based on domain decomposition. The model developed is applied to prediction of air quality in Tomsk for several chosen dates of 2000. The results of comparison of predictions and observations demonstrate applicability of the proposed approach, whose main advantage is high performance in forecasting of air pollution distribution over urban areas.