Markov processes, dynamic entropies and the statistical prediction of mesoscale weather regimes

Abstract

The data pertaining to transitions between weather regimes grouped in 3 clusters are analyzed. Evidence that the dynamics of transitions between the regimes is not a first order Markov process is presented on the basis of the properties of residence time distributions. This conclusion is corroborated further by an entropy analysis revealing that the process is characterized by long range temporal correlations. Simple models of this behavior are developed and the repercussions on the problem of prediction are discussed.