PhD Defense by Jade Omotoyosi Nina Brauns

Title: Fingerprints and conceptual models: reconstructing the behaviour of potential climate tipping systems

Abstract:

The study of the evolution of complex climatological phenomena can be hindered by limited observational data. In the absence of direct or indirect measurement, climate models can be employed to reconstruct the past and project the future behaviour of the climate. In this thesis, models on each end of the hierarchy are employed to study vital components of the climate system and climate phenomena, motivated by improving our understanding of abrupt transitions in the climate.
    
A minimal conceptual model for glacial cycles proposed by MacAyeal (1979) is revisited and expanded upon to investigate whether the core dynamical features of this paleoclimate phenomenon can be reproduced. The time asymmetry and change in periodicity of glacial cycles during the Mid-Pleistocene Transition are not explained by a purely astronomically forced system. Therefore, an additional multiplicative control parameter representative of an internal climate feature is included in the model, which results in a non-linear dynamical response of the system to the external solar forcing. A rule-based evolution of this secondary parameter is imposed, where this parameter controls the structural response to the solar forcing. The periodicity of the response is locked to that of the solar forcing, whereby the system can undergo a bifurcation during insolation maxima. A change in timescales of the secondary parameter can produce a change in periodicity of the response, while the abrupt shifts associated with glacial terminations are modelled as saddle-node bifurcations, allowing the model to replicate the time-asymmetry associated with glacial cycles.
                
Models across the hierarchy have provided evidence that the Atlantic Meridional Overturning Circulation (AMOC) and Subpolar Gyre (SPG) are potentially bistable systems that may, under projected future warming and freshwater forcing, transition to a weaker state. Analyses of the stability of these systems typically attempt to detect a loss of resilience in a fingerprint which acts as a proxy for the system. Therefore, the reliability of the stability analysis is dependent upon the accuracy of the fingerprint, motivating the need for robust AMOC and SPG fingerprints. 
    
Using a subset of CMIP6 models, novel fingerprints for AMOC and SPG are constructed using ridge regression applied to sea surface temperature and salinity fields. The ridge fingerprints are highly correlated with AMOC and SPG in pre-industrial control and historical simulations, with a larger spread in performance between models in projected forcing scenarios. This is likely a result of the non-stationary relationship between the surface ocean fields and the target system, under varying forcing. The ridge fingerprint is capable of reconstructing pre-industrial control and historical AMOC variability with zero lag, when contrasted with traditional sea surface temperature and salinity fingerprints. The ridge fingerprints are also capable of robustly reconstructing AMOC trends under various projected forcing levels, with lower performance in reconstructing SPG trends, perhaps as a result of the larger spatial scale of the AMOC when compared with the SPG. 

Chair of the assessment committee: Jens Hesselbjerg Christensen

Committee members: Peter Ashwin (University of Exeter), Bo Christiansen (DMI)

Principal supervisor: Peter Ditlevsen