Using of rank distributions in the study of perennial changes for monthly average temperatures

The possibility of comparing the climatic data of various years with using rank distributions is considered in this paper. As a climatic data, the annual variation of temperature on the spatial areas of meteorological observations with high variability in average temperatures is considered. The results of clustering of the monthly average temperatures values by means of a recurrent neural network were used as the basis of comparing. For a given space of weather observations the rank distribution of the clusters cardinality identified for each year of observation, is being constructed. The resulting rank distributions allow you to compare the spatial temperature distributions of various years. An experimental comparison for rank distributions of the annual variation of monthly average temperatures has confirmed the presence of scatter for various years, associated with different spatio-Temporal distribution of temperature. An experimental comparison of rank distributions revealed a difference in the integral annual variation of monthly average temperatures of various years for the Northern Hemisphere.