Abstract: Values of meteorological elements observed 4 times each day in Guangzhou were processed by Principal Component Analysis. Weather on every day was clustered into 5 weather types for winter and 6 for summer, according to the scores of 4 top components. Weather feature and Pattern of circulation of each weather type generally consist with existing experiences, but the procedure makes an objective classification. The structure of climate and dynamic features of weather types are revealed by the frequences of occurrence and transfer of weather types. Although the classification is based on routine observed meteorological variables, the weather types obtained well correspond to many phenomena in living things and environment. This shows that the procedure suggested and weather types created have great potential in studies on synoptic climatology and applied meteorology.