EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Predicting past tipping points: The Dansgaard-Oeschger events of the last glacial period

Johannes Lohmann and Peter Ditlevsen
Johannes Lohmann and Peter Ditlevsen
  • University of Copenhagen, Physics of Ice, Climate and Earth, Niels Bohr Institute, Copenhagen, Denmark (

The Dansgaard-Oeschger (DO) events of the last glacial period provide a unique example of large-scale climate change on centennial time scales. Despite significant progress in modeling DO-like transitions with realistic climate models, it is still unknown what ultimately drives these changes. It is an outstanding problem whether they are driven by a self-sustained oscillation of the earth system, or by stochastic perturbations in terms of freshwater discharges into the North Atlantic or extremes in atmospheric dynamics.

This work addresses the question of whether DO events fall into the realm of tipping points in the mathematical sense, either driven by an underlying bifurcation, noise or a rate-dependent instability, or whether they are a true and possibly chaotic oscillation. To do this, different ice core proxy data and empirical predictability can be used as a discriminator.

The complex temporal pattern of DO events has been investigated previously to suggest that the transitions in between cold (stadial) and warm (interstadial) phases are purely noise-induced and thus unpredictable. In contrast, evidence is presented that trends in proxy records of Greenland ice cores within the stadial and interstadial phases pre-determine the impending abrupt transitions and allow their prediction. As a result, they cannot be purely noise-induced.

The observed proxy trends manifest consistent reorganizations of the climate system at specific time scales, and can give some hints on the physical processes being involved. Nevertheless, the complex temporal pattern, i.e., what sets the highly variable and largely uncorrelated time scales of individual DO excursions remains to be explained.

How to cite: Lohmann, J. and Ditlevsen, P.: Predicting past tipping points: The Dansgaard-Oeschger events of the last glacial period, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-16665,, 2020

Display materials

Display file

Comments on the display material

AC: Author Comment | CC: Community Comment | Report abuse

Display material version 1 – uploaded on 30 Apr 2020
  • CC1: Comment on EGU2020-16665, John Bruun, 05 May 2020

    Dear Johannes and Peter

    The topic of these process being a spontaneous activation or linked to a more non-linear type of transport induced asymmetric oscilator is really interesting. 

    With the evidence in Lohmann (2019) suggesting that there is predictability in the pre-transition trajectory, do you think that we could explain this form of process using an non-linear oscillator phenomena - such a a recharge or delay oscillator non-linear type ? 

    I have been looking at this type of problem using Random Matrix Theory to assess the distinction between stochastic/noisy systems that a) have interacting eigenmodes and b) do not have interacting eigenmodes. We recently showed (Bruun and Evangelou, 2019) that such systems can be distinguised by assessing the nature of the extreme transport processes. I'll be sharing this work during this Tipping Points session.

    The consequences of our work are that stochastic systems with interacting eigenmodes have a dynamic robustness and the system themalizes. On the other hand if the system does not have interacting eigenmodes then the consequences are that the noise induced system would correspond to localized waves - where the extremes can be very heavy tailed.

    In your Fig. 2 of your slide, is their the possibility that as times approach 0 the density falls away from the exponential curve? That would be consistent with a Wigner surmise type setting where eigenmodes interact (and do not overlap each other) and perhaps some eigenmodes are losing thier interaction properties due to increased noise. 

    We can test this hypothesis by applying the characteristic ploynomial and extreme value process test (and proof) explained in our work. The GEV shape parameter can help us to diagnose what type of variability is being displayed in such abrupt transition systems. 

    Hope that helps - and it would be good to talk through this further with you, and I can explain the logic and how we test such systems. It would be very interesting to work with you on this.

    You can contact me at

    Best John