SIMULATED ANNEALING AND WEATHER REGIMES CLASSIFICATION

A new approach based on the travelling salesman problem and simulated annealing is used to classify weather maps and to associate them with weather regimes. The usual classification methods are able to identify the preferred patterns of large-scale atmospheric flow. They fail, ho wever, in characterizing the boundaries of various regimes which are g enerally a result of the geometrical constraints imposed by the chosen method. The method is aimed at overcoming this difficulty and helping to define the boundaries between weather regimes in order to study th e characteristics of and transitions between these regimes. Weather re gimes are derived from a series of 37 winters of 700 hPa geopotential height observations using the algorithm developed by Vautard (1990). T he new approach is then applied to assign each observation to its corr esponding weather regime. It is shown that the method provides results which agree with the literature and helps to establish recurrence in agreement with observations. We then, speculate on the relationship wh ich may link the boundaries of weather regimes to the transitions betw een these regimes.