Climate networks as a tool for data-driven hypothesis generation

Abstract

Over the past decade, climate networks have emerged as a powerful tool to characterise high dimensional weather and climate datasets. Climate networks are a sparse representation of the dynamical similarities between weather time series from different geographical locations. Nodes represent the locations themselves, and network edges represent high dynamical similarity between pairs of locations. The topology of the resulting complex network encodes information about how atmospheric and oceanic dynamics “connect” different locations. For instance, strong monsoon years might yield a different network structure than weak monsoon years. With the tools of graph theory and complex networks at our disposal, we can characterise climate dynamics in novel and interesting ways, which yield, in part, results that corroborate what meteorologists already know, and, in part, results that generate new hypotheses about how atmospheric and oceanic processes influence different weather patterns. In this...