# WP5: Tipping Points in the Climate System

An important topic in which the innovative introduction of network thinking in climatology can have important rewards is in the characterization of regime transitions in the climate systems. Methodologies to detect regime shifts in past time series have been developed, both in the climatic context and beyond. They usually consider the time evolution of local quantities and search for classical indicators of temporal slowing down or enhanced fluctuations. In some cases, data transformations (such as principal component analysis) that potentially combine information from many locations are also used. Here we propose to develop novel techniques to directly analyze the correlations and interactions between different locations, as represented by climatic networks. We believe that the sensitivity of such approaches will be higher than of the classical ones. GCM simulations and time series of well studied observed climatic shifts will be used to develop the new

methodology, and output from modelling exercises such as CMIP5 will be also searched for predicting the approach to future tipping points.

Specifically objectives of WP5 are:

1) to develop theoretically (e.g. by means of percolation theory) classification schemes which could characterize topological phase transitions in networks;

2) to analyze the time evolution of the network indicators developed here, and in WP1 and WP4, identifying from them useful topological changes which can be used to locate and to predict climate regime changes. Particular attention will be given to well-studied phenomena such as the MOC collapse or the different ENSO phases;

3) to compare the power of these new indicators with classical ones of bifurcation behaviour such as divergent variances or correlation times and

4) to evaluate the suitability of Lagrangian networks to identify tipping points, e.g. by focussing on singularities of Jacobian or transilience matrices.