Master Thesis Defense by Alicja Kalucka

Title: Reservoir computer-based detection of AMOC tipping

Abstract:
The undeniable role of the Atlantic Meridional Overturning Circulation (AMOC) in the global climate system is recognized. The potential consequences of stopping or reversing the AMOC are of great concern. The aim of this study is to assess whether a tipping point in the AMOC could be forecast using modern machine learning methods.
An idealized coupled ocean-atmosphere model was used to train a reservoir computer, into which a bifurcation channel was incorporated.
This channel, which can identify complex system behaviour, allows critical transitions to be detected and system collapse to be forecast.
The majority of the predictions made by our reservoir computer were found to align with the outputs of our model, with most values falling within three standard deviations. These findings suggest that collapses in low-order climate models might potentially be forecast using this machine-learning approach. This could lead to an enhanced understanding of abrupt climate changes and the ability to predict them.

Supervisors: Peter Ditlevsen and Henk Dijkstra   
Censor: Jens Olaf Pepke Petersen