E.V. Kadantsev, E.V. Dmitriev, A.I. Chavro

At present, general circulation models describe very accurately the state of the atmosphere on a global scale. At the same time on the regional scale results obtained in model experiments were not sufficiently realistic. Since the demand of consumers of weather forecasts and climate information about the detailed spatial structure of meteorological fields can not be satisfied by the interpolation approach, the problem of the local interpretation of large-scale prediction, related to a class of ill-posed inverse problems, which received the name of «downscaling problem» in the western publications, arose. In this article we propose a statistical model for solving the inverse problem of recovering the regional structure of the field of average temperature in the Alpine region using predicted large-scale field of this value. The problem is solved by the me-thod best in the mean-square estimation of the desired vector, using a priori information in the form of first and second moments of the predictor and predicted value. Numerical experiments have shown that despite the selection of region of complex topography, large temperature difference in time and space, the model allowed to recover the average temperature in the network of meteorological stations with an accuracy of 1,5°C with the natural variability of 3,2°C. In particular, for the region with the coordinates 9°?12° el, 44°?46,5° nl. it managed to recover about 65% of variability for the fields of average monthly temperature, which is quite acceptable results.