Reliability of Inference of Directed Climate Networks Using Conditional Mutual Information

Abstract

Across geosciences, many investigated phenomena relate to specific complex systems consisting of intricately intertwined interacting subsystems. Such dynamical complex systems can be represented by a directed graph, where each link denotes an existence of a causal relation, or information exchange between the nodes. For geophysical systems such as global climate, these relations are commonly not theoretically known but estimated from recorded data using causality analysis methods. These include bivariate nonlinear methods based on information theory and their linear counterpart. The trade-off between the valuable sensitivity of nonlinear methods to more general interactions and the potentially higher numerical reliability of linear methods may affect inference regarding structure and variability of climate networks. We investigate the reliability of directed climate networks detected by selected methods and parameter settings, using a stationarized model of dimensionality-reduced surface air temperature data from reanalysis of 60-year global climate records. Overall, all studied bivariate causality methods provided reproducible estimates of climate causality networks, with the linear approximation showing Entropy 2013, 15 2024 higher reliability than the investigated nonlinear methods. On the example dataset, optimizing the investigated nonlinear methods with respect to reliability increased the similarity of the detected networks to their linear counterparts, supporting the particular hypothesis of the near-linearity of the surface air temperature reanalysis data.

Regional and inter-regional effects in evolving climate networks

“…For other issues related to the choice of an appropriate dependence measure see e.g. Paluš et al (2011) and Hlinka et al (2013b).…”

“…Recently, Vejmelka et al (2014) proposed a principled approach for estimation of the count of dynamically relevant components in a climate time series data set and their identification (a basic description of the method is available already in Hlinka et al (2013b), where it was used along with gridbased dimensionality reduction to assess the effects of nonlinearity for causal network estimates). We use this method in conjunction with varimax-rotated principal component analysis (Vejmelka and Paluš, 2010;Groth and Ghil, 2011) as an intermediate step in network analysis, as a method for concise representation of the data compared to the (climatologically largely irrelevant) spatial sampling used in the original gridded data set.…”

Unified functional network and nonlinear time series analysis for complex systems science: Thepyunicornpackage

“…While lagged MI can be used to quantify whether information in Y has already been present in the past of X, this information could also stem from the common past of both processes and, therefore, is not necessarily a sign of a transfer of unique information from X to Y. A first step towards a notion of directionality (the more demanding causality problem is discussed at the end of this section) is to assess a bivariate notion of information transfer between two time series 79,80 in order to exclude this common past. Here, we consider two measures to achieve this goal, implemented in CouplingAnalysis.…”

Distinguishing the effects of internal and forced atmospheric variability in climate networks

“…and the reliability and robustness of the networks uncovered have also been analyzed in terms of a critical comparison of the networks found with the various methods used (Paluš et al, 2011;Hlinka et al, 2013;Martin et al, 2013;Tirabassi and Masoller, 2013). A main conclusion of these studies is that it is crucial to analyze the robustness of the method used to quantify climate similarities because trends and serial correlations in the time series, as well as time lags, can significantly affect the topology of the network obtained.…”