PhD defence: Methods for the Study of Rare Transitions in the Atlantic Meridional Overturning Circulation

PLEASE NOTE: The candidate gives a layman's talk, therefore the livestream will start fifteen minutes earlier.

The Atlantic Meridional Overturning Circulation (AMOC) plays a capital role in the transport of heat and salt across the Atlantic Ocean towards the Northern Hemisphere. However, global warming may trigger its fast disruption, which would have a great impact on the climate all over the world. It is therefore capital to determine the likelihood of such an event, and study its mechanisms and consequences to make sure that society is prepared.

However, this task requires the computation of costly model simulations, which is usually too expensive to be done in state-of-the-art climate models. Instead, we study in this thesis algorithms that facilitate the simulation of rare and extreme events (such as an AMOC collapse), by decreasing the associated costs. But running such algorithms remains computationally expensive, so we must maximise their efficiency and reliability before applying them on large climate models.

One way to optimise them, explored in the first half of this thesis, is to improve the measure of the closeness of a given simulation to an AMOC collapse. We achieve this task by setting up a machine learning scheme, while trying to use as little training data as possible (since data is scarce when working with large climate models). Another interesting approach, explored in the second half of this thesis, is to extract as much information as possible from the algorithm's output, to analyse and track the drivers and impacts of such events.