Cluster analysis for pattern recognition in solar butterfly diagrams

We investigate to what extent the wings of solar butterfly diagrams can be separated without an explicit usage of Hale's polarity law as well as the location of the solar equator. We apply two algorithms of cluster analysis for this purpose, namely DBSCAN and C-means, and demonstrate their ability to separate the wings of contemporary butterfly diagrams based on the sunspot group density in the diagram only. Then we apply the method to historical data concerning the solar activity in the 18th century (Staudacher data). The method separates the two wings for Cycle 2, but fails to separate them for Cycle 1. In our opinion, this finding supports the interpretation of the Staudacher data as an indication of the unusual nature of the solar cycle in the 18th century. (C) 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim