Use of Temporal Principal Components Analysis to Determine Seasonal Periods

Abstract

Temporal principal components analysis was applied separately to monthly long-term wind, temperature, and precipitation data for Southern California. Physical explanations of the significant eigenvectors are presented. Cluster analysis of the component loadings was then used to form groups of months (seasons) having similar spatial patterns. The resulting groupings of months differed from the conventional definition of seasons. The wind and temperature analyses grouped the same months, with long summers, moderately long winters, short springs, and very short autumns. The precipitation analysis formed a long season, including the winter months, representing synoptic systems occasionally passing through the area, a summer thunderstorm season associated with influx of moisture from the south, and dry transitional periods separating these seasons. The purpose of the analysis was to pregroup two years of hourly wind data to remove most of the annual signal before applying spatial eigenvector analysis for a mesoscale climatological classification study. The approach is expected to be most useful when applied to mesoscale areas with significant seasonal variation in spatial patterns of climatic variables.