A new method of cluster analysis for numerical classification of climate

Abstract

In this paper, a new method of cluster analysis incorporating a statistical test, was proposed in order to decide whether to accept or reject the assumption of a numerical classification of climate according to aχ 2 test. On the basis of normality of errors, theχ 2 test has been derived from the variance of the Euclidean distance or from that of the reduced distance. This new method of cluster analysis consists of uniform clustering and central clustering. The former is adapted to a more detailed numerical classification of climate, while the latter is appropriate for the large-scale numerical classification of climate. Examples of calculation and application of this new method of cluster analysis are presented in order to illustrate calculations undertaken in the processes of climate classification.