Defining climatological seasons using radially constrained clustering

Abstract

A general method for defining seasons is developed that can be applied to any set of climatological variables. Radially constrained clustering is based on stratigraphically constrained clustering, which is a form of cluster analysis that imposes constraints on the temporal contiguity of the solutions. Time is represented cyclically rather than linearly and the climatological seasons are allowed to cross the calendar year boundary. To demonstrate, the method is applied to hydroclimatological data from British Columbia, Canada. The proposed clustering method is capable of separating the data into meaningful seasons. Trends in the start/end dates and lengths of the seasons are consistent with those identified in previous research. Copyright 2005 by the American Geophysical Union.