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

The statistical model for reconstruction of small-scale meteorological fields, in particular, the fields of surface temperature (average, minimum and maximum) based on the large-scale one in the region of English Channel (La Manche) is considered. Time series of surface temperature extracted from NCAR / NCEP reanalysis, given in a grid with a resolution of 2.5°×2.5°, are used as predictors. Small-scale field was presented by observations of surface air temperature at meteorological stations in the northern coast of France, the region of English Channel. Task was complicated by a large amount of gaps in the observational data. Numerical experiments showed that the model allows to recover about 80% of the natural variability (the error of about 1.3°C) of the average temperature field and about 75% of the variability of the extreme temperature fields (the error of about 1.5°C and 1.9°C for the minimum and maximum temperatures respectively). As expected, the procedure of filling gaps markedly increased the accuracy of reconstruction. Absolute error of the solution was decreased for about 30-35%. Analysis of correlations between reliability of the model and accuracy of the inverse problem solution showed that the reliability parameter can be with some confidence level used for the allocation of realizations to have large reconstruction error, or extreme situations in the atmosphere.