Technique of ultra-short-term forecast of atmospheric parameters based on the Kalman filtering algorithm and 2D dynamic-stochastic model of the regression type is considered, and the results of its qualitative estimate for the boundary layer (up to the altitude of 800 m) based on meteorological and aerological observations are discussed.