Risk minimization for atmospheric data restoration

The paper deals with filling gaps in meteorological fields aimed at weather simulation. A universal local approximation approach is described which allows us to fill gaps in meteorological fields under various conditions, and at the same time, effect time-step weather forecasts. We present results of small sample gap filling for some meteorological elements. The results are compared against those obtained with local averaging and inertia methods. Some possibilities of data restoration and prediction based on available global meteorological data are discussed.