Smooth information flow in temperature climate network reflects mass transport

Jaroslav Hlinka, Nikola Jajcay, David Hartman, Milan Paluš
2017 Chaos  
A directed climate network is constructed by Granger causality analysis of air temperature time series from a regular grid covering the whole Earth. Using winner-takes-all network thresholding approach, a structure of a smooth information flow is revealed, hidden to previous studies. The relevance of this observation is confirmed by comparison with the air mass transfer defined by the wind field. Their close relation illustrates that although the information transferred due to the causal
more » ... ce is not a physical quantity, the information transfer is tied to the transfer of mass and energy. Recently the complex network approach to analysing dynamical systems reached also the climate science and became one of the tools for uncovering dependence structures and teleconnections in the atmospheric data. Usually, for the sake of clarity and interpretability, symmetric statistical measures such as correlation or mutual information are taken into account. However, the drawback is that these measures lack the notion of directionality. In this paper we investigate the possibility of inferring the causal climate network with the explicit direction of causal influence, taking the conservative approach of linear Granger causality applied to gridded temperature data. Following this avenue, in conjunction with a novel winner-takes-all thresholding scheme, yields an easy-to-interpret causal network with smooth information flow structure. To assess the significance of this observation, we designed random graph models for climate networks and quantitatively compared the temperature causal network with prevailing wind direction.
doi:10.1063/1.4978028 pmid:28364752 fatcat:qffnlnpmvzho5ecbhe52mbdbum