Using Networks to Understand Climate Dynamics
Networks pervade nature: From understanding hyena behavior to modeling earthquake physics, a range of complex phenomena can be modeled using basic concepts of graph theory and network phenomena.
One of these phenomena is central to our lives: precipitation patterns. A study published in PLoS ONE used network theory to model patterns in precipitation on large geographical scales. This is part of a larger field of study, known as climate networks.
As is common in these types of studies, the first step in modeling network phenomena in complex natural systems is to parametrize them by identifying nodes and edges. In the study under consideration, the nodes in the network were geographical regions of fixed data. To form an edge between nodes, the cross-correlation function (essentially the result of taking the integral of the inner product of two time-series) was evaluated between all nodes to understand how similar the measurements between them were- if they were similar or dissimilar enough between two nodes, an edge was generated. This method resulted in there being 1674 nodes used in the study.
Once a network (consisting of nodes and edges between nodes, with no self-loops) was generated, we can generate several parameters characterizing the network. One of these in the clustering coefficient for a node, describing the probability that two neighbors of a node are also neighbors. The clustering coefficient showed notable heterogeneities in specific regions, such as in Africa, East Asia, and South America.
By analyzing the network, the authors of the study came to some interesting conclusions. Firstly, the network was shown to be made of multiple connected components of a range of size. Of the 1674 nodes, 1632 nodes were linked and formed a subnetwork, whilst the other 14 nodes formed smaller networks.
More interestingly, cross-correlations showed that links were present that could be linked to the presence of the Gulf Stream, a large oceanic current. In addition, extreme events were shown to simultaneously occur worldwide, indicating that planetary-scale waves were propagating throughout the Earth. In addition, the node degree distribution clearly indicated a scaling-law behavior, similar to the power-law type relationships studied during class.
Overall, the study clearly indicated that analyzing climatic systems using complex network principles presents strong potential to develop a greater understanding of one of the world’s most complex natural systems: Climate Dynamics.
Scarsoglio, S., Laio, F., & Ridolfi, L. (2013). Climate Dynamics: A Network-Based Approach for the Analysis of Global Precipitation. PLoS ONE, 8(8). doi:10.1371/journal.pone.0071129