Markov chain analysis of regional climates

Abstract. We present a novel method for regional climate classification that is based on coarse-grained categorical representations of multivariate climate anomalies and a subsequent Markov chain analysis. From the estimated transition matrix several descriptors, such as persistence, recurrence time and entropy, are derived. These descriptors characterise dynamic properties of regional climate anomalies and are connected with fundamental concepts from nonlinear physics like residence times, relaxation process and predictability. Such characteristics are useful for a comparative analysis of different climate regions and, in the context of global climate change, for a regime shift analysis.

We apply the method to the bivariate set of water vapour and temperature anomalies of two regional climates, the Iberian Peninsula and the islands of Hawaii in the central Pacific Ocean. Through the Markov chain analysis and via the derived descriptors we find significant differences between the two climate regions. Since anomalies are departures from seasonal and long term components, these differences relate to differences in the short term stability of both regional climates.