Temperature distributions comparison by the clustering of their proximity measure

In this paper the method for determining of proximity of the temperature measurements fields is described. The method includes a clustering of values of the measurements fields, an estimation of their informational proximity to the chosen reference field by Kullback-Leibler distances and a clustering of the values of these distances. Clustering is performed by recurrent neural network. The results of applying the proposed method to the analysis of temperature fields in the Northern hemisphere for the period of 1955-2010 are presented.