Pattern of climate network blinking links follows El Niño events

Using measurements of atmospheric temperatures, we create a weighted network in different regions on the globe. The weight of each link is composed of two numbers —the correlations strength between the two places and the time delay between them. A characterization of the different typical links that exist is presented. A surprising outcome of the analysis is a new dynamical quantity of link blinking that seems to be sensitive especially to El Niño even in geographical regimes outside the Pacific Ocean. Copyright c © EPLA, 2008 Introduction. – El Niño, a massive burst of heat exchange between the ocean and the atmosphere in the Pacific Ocean, is regularly tracked by measurements of sea surface temperatures and pressure differences in that zone [1]. When these measures reach a certain threshold level, the event begins. The end of the event is defined to be the time when the temperatures and pressures reach the threshold level again on their way to their base line [2]. This procedure cannot hold for other geographical zones that are also influenced. The reason is that in these zones the temperature fluctuations due to El Niño are of the same order as other fluctuations of temperature, due to other dynamical processes not related to El Niño. Another reason is that the pattern of an El-Niño–forced atmosphere is quite complex, yielding different qualitative and quantitative influences in different places (see, e.g., [3]). Here we show that by observing the pattern of climate network links we can follow El Niño events even in zones where the temperature does not show any effect. The climate network is composed of nodes and links. The nodes are confined geographical zones that correspond to a single point of measurement on a grid. Each node carries a measured state variable that changes in time. The state variable in the current study is the temperature time series. A link between two nodes exists if there is a significant cross-correlation between their state variables’ time series. A high peak in the cross-correlation function represents a strong link, and its position in time represents the time delay between the two nodes. (a)E-mail: [email protected] The distribution of the time delays is qualitatively similar everywhere around the globe. In every geographical zone studied there is always a small fraction of links between places whose temperatures oscillate coherently with very short time delays. There is a second, larger collection of pairs that exhibit intermediate correlation and change their time delays frequently. The third group is the group of pairs that show weak correlation. Our results suggest that mainly the last two groups modify their behavior in the presence of El Niño events (see fig. 1). Using these general observations, we develop our analysis further by applying a symbolization approach to the time series composed of the cross-correlation values. The analysis exploits the highly increased sensitivity of correlation fluctuations to El Niño events, thus introducing a way to keep track of El Niño influences around the world. Recent studies on climate networks [4,5] have identified differences in the climate network structure between El Niño and non–El-Niño time periods. Yamasaki et al. observed structural changes in the network. During El Niño many links disappear. Tsonis and Swanson divided their records into three parts —accumulation of all El Niño epochs, accumulation of all La Niña epochs and accumulation of the residual parts, and built a climate network based on temperature correlations from each part separately. They found differences in spatial distribution of links, yielding, on the whole, significantly less links in the El-Niño–based network. Our present study strongly supports refs. [4,5], adding detailed timedependent information —the blinking of links during El Niño, which was missing.