Defining an early warning method for an AMOC collapse based on ensemble statistics

The Atlantic Meridional Overturning Circulation (AMOC) is a crucial climate component, as its potential collapse would constitute a significant response to Earth’s changing climate. This critical transition has been the subject of numerous studies over the years, both from the aspect of climate modeling and dynamical systems theory. In the context of the latter, climate change is a process in which a complex, chaotic-like system possesses time-dependent parameters, in the form of e.g. the growing CO2 concentration. It has been known that such systems have a chaotic attractor which is also time-dependent, a so-called snapshot attractor. Such objects, and thus the systems they describe, can only be faithfully represented by a probability distribution over an ensemble of simulations, so-called parallel climate realizations.

Based on this probability distribution, we define a novel early warning indicator for crucial transitions such as an AMOC collapse. The AMOC is said to possess a multistable quasipotential landscape, and the collapse is a transition between stable states. We argue that, from the point of view of statistical physics, this is analogous to a phase transition, but in a non-adiabatic setting. As such, the variance of the distribution over the ensemble is expected to develop a local maximum around the transition point, giving rise to a potential early warning by identifying the preceding maximum of its derivative. This method is first demonstrated on a conceptual climate model, before the analysis is carried out on ensemble simulations from the ACCESS-ESM model. The analysis in the former case is simpler, while in the latter, one has to contend with the dependence of the AMOC strength on spatial coordinates, resulting in multiple early warning points for different depths and latitudes.