Dansgaard-Oeschger Events as Laboratory for Extremes Variability under AMOC Collapse

Obtaining a statistical description of the extreme events that occur in a system that exhibits tipping behaviour is challenging due to the strong changes that the system undergoes. In this work, we use non-stationary Generalized Extreme Value (GEV) distributions to study the statistics of the extremes while capturing their temporal variability relative to a covariate data series which can be a driver or a response to the tipping. We exemplify this methodology by employing 8000 year long CCSM4 simulations with low concentrations of atmospheric CO₂ that show spontaneous D-O oscillations. This setting allows to study the minimum annual temperatures across the globe as a function of the temporal variability of the strength of the AMOC. The parameters of the distribution convey information about how the nature of the changes observed and its spatial variability, giving an insight on how the strength of the AMOC is related with the magnitude, variability and tails of the distributions. The extrapolation capabilities of this method are discussed compared to other studies and mechanisms of AMOC collapse.