Early warning signals in different AMOC models

Abstract

In this thesis, the main goal is to find mathematical measures which can function as early warning signals for the Atlantic Meridional Overturning Ciculation (AMOC). Because of climate change, the earth is warming and if this continuous, there is a probability that the AMOC tips. It is interesting and helpful to find early warning signals which can hopefully predict this tipping. Hence, we first introduce an AMOC model, namely the Cessi model. Next, we can use different solution methods to solve that model. We define a Schr¨odinger approach, we use a Monte-Carlo approximation and we try to improve this Monte- Carlo approximation with help of two different kinds of Neural Networks. For the non-dimensional Cessi model, we find that the Schr¨odinger approach works the best. We can use this solution to calculate different possible early-warning signals. The eigenvalues, the Entropy and the Probability Current seems to give the best early-warning signals. To check those early warning signals, we can compare that for more detailed models. However, most of them are hard to solve with the Schr¨odinger solution method. The Monte-Carlo approximation is possible and it can be improved with the neural networks. Nevertheless, this is a topic of further research.