Topology and Predictability of El Niño and La Niña Networks

We construct the networks of the surface temperature field for El Niño and for La Niña years and investigate their structure. We find that the El Niño network possesses significantly fewer links and lower clustering coefficient and characteristic path length than the La Niña network, which indicates that the former network is less communicative and less stable than the latter. We conjecture that because of this, predictability of temperature should decrease during El Niño years. Here we verify that indeed during El Niño years predictability is lower compared to La Niña years.

Figure 1

Total number of links at each geographic location (grid point) as delineated by the fraction of the total global area that a point is connected to. The top panel corresponds to El Niño years and the bottom panel to La Niña years. Both networks are characterized by a presence of supernodes (nodes with many links) albeit less pronounced in the El Niño network.

Figure 2

Degree distribution for the La Niña network (solid line) and the El Niño network (broken line). For most of the scales involved, the distribution is a power law.

Figure 3

The relative frequency distribution of links as a function of link length. Clearly during El Niño events the number of links at all space scales is reduced.

Figure 4

Monthly averaged variance of the surface temperature for the tropics ($20°\mathrm{S}–20°\mathrm{N}$) from persistence for 10 day lead forecasts at a given location. Here data in the period 1979–2006 are used. Prior to that tropical daily temperature fields are less reliable due to poor satellite coverage. As explained in the text, this figure suggests that for El Niño events the variance is increased compared to La Niña events indicating loss of predictability during El Niño events.