Skip to content


Disentangling climate interactions and inferring tipping points by using complex networks

Giulio Tirabassi, Presentation date: June 5, 2015

Author: Giulio Tirabassi
Title: Disentangling climate interactions and inferring tipping points by using complex networks
Director: C. Masoller
Presentation date: June 5, 2015
Link to text:

Abstract: In a scenario of major global climatic changes, the understanding of the Earth system has become in recent years an impelling task of the scientific community. In the last decade, an increased performance of supercomputers as well as a better understanding of the physical processes underlying Earth dynamics (such as cloud formation, moist exchange between soil and atmosphere, and so on) improved dramatically the quality of climate models. Moreover, an increasing satellite coverage allowed the generation of very detailed databases, called reanalysis data, that give with good precision the state of the most important dynamical variables at high spatio-temporal resolution for at least the past 40 years. This huge amount of data is being used by the climatological scientific community to investigate the nature of the ongoing physical processes in the Earth system. The increasing of such datasets motivates the development of new analysis. In particular, since in first approximation the climatological behaviour of the atmosphere can be described by relatively simply linear models, the non-linear data analysis has been largely overlooked so far. However, many relevant climatic processes have a strong non-linear component. A paradigmatic example is the so-called El Niño-Southern Oscillation, a coupled ocean-atmosphere process that can be sketched in first approximation as a non-linear oscillator with delayed feedback. Another phenomenon of the Earth system in which the non-linearities play a major role is doubtlessly the barotropic polar jet, that is maintained by the non-linear stress produced by its own eddies. In this way a complex system of positive and negative feedbacks is established, influencing the dynamics of the major actor in the synoptical variability in the high latitudes. Given these considerations, it is important to study such phenomena with tools coming from the field of complex systems. In particular, in this thesis we present new data analysis techniques based on information theory and complex networks that take into account the non-linear nature of the climate processes under examination, as well as give a new perspective to climatological data analysis. Complex networks have emerged in the recent years as a useful and powerful tool to investigate a large variety of phenomena in which it is possible to identify a discrete number of components among which relations can be established. Being this a very general concept, it is not surprising that it found successful application in very different field such as sociology, biology, epidemics, geophysics, and so on. In particular in this thesis we will focus in the construction of what are called climate networks, i. e. networks in which the nodes are composed by geographical locations on Earth and the links are given by relations among them. We will generally consider links computed from the statistical similarity of the dynamics of a climatological variable available at each geographical location, such as the surface air temperature (SAT). This is a quite general approach that, depending on the definition of the statistical similarity employed, allows to investigate different characteristics of the variable under examination. We will define the statistical similarity using concepts that are adopted from information theory.