Topological computation analysis of meteorological time-series data

Abstract

A topological computation method, called the MGSTD method, is applied to time-series data obtained from meteorological measurement. The method gives decomposition of the dynamics into invariant sets and gradient-like transitions between them, by dividing the phase space into grids and representing the time-series as a combinatorial multi-valued map over the grids. Since the time-series is highly stochastic, the multi-valued map is statistically determined by taking preferable transitions between the grids into account. The time-series data are principal components of pressure pattern in troposphere and stratosphere in the northern hemisphere. The application yields some particular transitions between invariant sets, which leads to circular motion on the phase space spanned by the principal components. The Morse sets and the circular motion are consistent with the characteristic pressure patterns and the change between them that have been shown in preceding meteorological studies.