Geophysical Research Abstracts, pp. EGU2014-8893, 2014.
Among other things, new methods are being developed at PIK that, on the one hand, investigate new aspects in palaeoclimate data, but also cope with the difficulties usually associated with palaeoclimate analyses – such as gaps in the data, age uncertainties, or irregularities in the data sampling. While this is basic research, it is also immediately applied to interesting research questions.
One focus is the development and application of methods that specifically investigate recurring patterns. These methods are very successful for finding abrupt changes, but also for comparing different data sets, e.g., to detect mutual influences.
In one of our first analyses of the three stalagmites BB-1 to BB-3, we had looked at the oxygen isotopes of BB-1 with recurrence based methods. Thereby we had explicitly considered dating uncertainties. Due to the dating uncertainties there are small shifts of the time series along the x-axis. Therefore, within the uncertainties, different realizations of the course of the measured oxygen isotopes are possible.
The recurrence patterns are examined with a special analysis tool, the so-called “recurrence plot”. This can be used to plot times when similar states have occurred. The patterns seen in such a recurrence plot have a deeper meaning and can be further analyzed and quantified to find changes in the dynamics of the underlying process.
The numerical analysis of recurrence plots yields various results that shed light on different aspects of climate dynamics. Two such results are briefly presented here: on the one hand, the complexity of climate variability (“transitivityTransitivity Ein Maß aus der Netzwerktheorie, welches quantifiziert, wie stark sich Zustände zu kleinen Gruppen zusammenfinden.”) and, on the other hand, how well such a climate signal would be predictable (“determinismDeterminism Ein Maß aus der Wiederkehr-Analyse, welches beschreibt, wie gut sich die Veränderung eines Systems vorhersagen läßt.”). Both results show a general tendency towards greater complexity and lower predictability for younger ages. However, they also show short-lived increases to better predictability for certain time points, namely about 4,200, 2,800, and 1,400 years ago. It was precisely at these times that short-lived and rapid glaciations occurred in the North Atlantic, the so-called “Bond events.” Moreover, there are variations in the complexity of the climate signal (“transitivity”), but these variations are within the range of uncertainty (confidence interval), so we cannot really interpret them.