An interaction network perspective on the relation between patterns of sea surface temperature variability and global mean surface temperature

On interannual- to multidecadal timescales variability in sea surface temperature appears to be organized in large-scale spatiotemporal patterns. In this paper, we investigate these patterns by studying the community structure of interaction networks constructed from sea surface temperature observations. Much of the community structure can be interpreted using known dominant patterns of variability, such as the El Nino/Southern Oscillation and the Atlantic Multidecadal Oscillation. The community detection method allows us to bypass some shortcomings of Empirical Orthogonal Function analysis or composite analysis and can provide additional information with respect to these classical analysis tools. In addition, the study of the relationship between the communities and indices of global surface temperature shows that, while El Nino-Southern Oscillation is most dominant on interannual timescales, the Indian West Pacific and North Atlantic may also play a key role on decadal timescales. Finally, we show that the comparison of the community structure from simulations and observations can help detect model biases.