Application of fuzzy sets to climatic classification

Abstract The theory of fuzzy sets is discussed as a possible way of dealing with the continuity of climatic data. The method of fuzzy c -means is described in detail and used to create fuzzy groups for two sets of climatic data, one from Australia and the other from China. The resulting groups seem intuitively reasonable, showing the inherent continuity of the data and a reasonable geographical contiguity. The problem of choosing c (the number of groups) and m (the degree of fuzziness) simultaneously remains unsolved. Measures of fuzziness, which have been used by some authors to estimate c given m , give rise to heuristic arguments which ignore the question of “what is a group?”. Some authors have assumed that large jumps in membership functions indicate levels at which natural groupings occur in data. An attempt experimentally to gain some agreement between standard measures of fuzziness and this approach, which agrees with that expected by intuition in a simple example, has led to a cursory yet promising approach to this problem. The fuzzy sets approach is realistic and flexible, and may offer a better approach to information transfer than does the classification of climate into discrete sets. The initial problems of accepting the notion of formal fuzzy classification and of perception of fuzzy classes, once established, appear from the authors' experience to be transient.