The Atlantic meridional overturning circulation (AMOC) is a large, basin-scale circulation located in the Atlantic Ocean that transports climatically important quantities of heat northward. It can be described schematically as a northward flow in the warm upper ocean and a southward return flow at depth in much colder water. The heat capacity of a layer of 2 m of seawater is equivalent to that of the entire atmosphere; therefore, ocean heat content dominates Earth’s energy storage. For this reason and because of the AMOC’s typically slow decadal variations, the AMOC regulates North Atlantic climate and contributes to the relatively mild climate of Europe. Hence, predicting AMOC variations is crucial for predicting climate variations in regions bordering the North Atlantic. Similar to weather predictions, climate predictions are based on numerical simulations of the climate system. However, providing accurate predictions on such long timescales is far from straightforward. Even in a perfect model approach, where biases between numerical models and reality are ignored, the chaotic nature of AMOC variability (i.e., high sensitivity to initial conditions) is a significant source of uncertainty, limiting its accurate prediction.
Predictability studies focus on factors determining our ability to predict the AMOC rather than actual predictions. To this end, processes affecting AMOC predictability can be separated into two categories: processes acting as a source of predictability (periodic harmonic oscillations, for instance) and processes acting as a source of uncertainty (small errors that grow and significantly modify the outcome of numerical simulations). To understand the former category, harmonic modes of variability or precursors of AMOC variations are identified. On the other hand, in a perfect model approach, the sources of uncertainty are characterized by the spread of numerical simulations differentiated by the application of small differences to their initial conditions. Two alternative and complementary frameworks have arisen to investigate this spread. The pragmatic framework corresponds to performing an ensemble of simulations, by imposing a randomly chosen small error on the initial conditions of individual simulations. This allows a probabilistic approach and to statistically characterize the importance of the initial condition by evaluating the spread of the ensemble. The theoretical framework uses stability analysis to identify small perturbations to the initial conditions, which are conducive to significant disruption of the AMOC.
Beyond these difficulties in assessing the predictability, decadal prediction systems have been developed and tested through a range of hindcasts. The inherent difficulties of operational forecasts span from developing efficient initialization methods to setting accurate radiative forcing to correcting for model drift and bias, all these improvements being estimated and validated through a range of specifically designed skill metrics.
Florian Sévellec and Bablu Sinha
Subject:Climate Systems and Climate Dynamics, Modeling