Weather categorization in climate change research: Preliminary results

Summary

The results of an objective weather categorization are presented. The 9 meteorological variables recorded daily during winter seasons of 1961–66 at Prague-Clementinum represent the input dataset. The principal component analysis and a few clustering procedures have been evaluated. 5 component solution and the average linkage clustering method were found optimal. The winter days have been grouped, according to their meteorological character, into 14 clusters. The warm categories represent 20% of the time and the cold categories less than 15% of the days. The mean maps of 1000 hPa and 500 hPa are shown for a few selected categories. Clustering techniques applied to long-time instrumental series can provide a better basis for attempting to detect temperature changes which have taken place over a long time span.