Variability of the Global Overturning Circulation

The oceanic global overturning circulation is critical for the distribution of temperature, salt, and carbon. It consists of warm ocean surface flows and colder flows in the deep and bottom layers of the ocean. These flows at different depths are connected through deep-convection and upwelling. The global scale of the overturning circulation implies that local anomalies, especially those at the convection sites, can have large-scale impacts, affecting also other ocean basins and the atmosphere. In this thesis the variability of the global overturning circulation is investigated with an emphasis on the only two regions where major deep-convection occurs: the North Atlantic and the Southern Ocean. In the North Atlantic, deep-convection is part of the Atlantic Meridional Overturning Circulation (AMOC). Model simulations suggest that the AMOC will weaken during the 21st century due to anthropogenic greenhouse gas emissions. However, the amplitude of the weakening is highly uncertain. In the first part of this thesis, the sources of uncertainty for future projections of the AMOC are identified. The analysis of models from the Coupled Model Intercomparison Project phase 3 and phase 5 (CMIP3 and CMIP5) reveals that the disagreement of the models, i.e., the model uncertainty, is the major contributor to the total uncertainty. Uncertainty due to internal variability is important only during the first few decades of the 21st century projection. Uncertainty caused by the possible range of future greenhouse gas emissions is negligible. The cause for the high model uncertainty in the AMOC projection seems to be that also the projections of the freshwater flux and of the gyre circulation are subject to high model uncertainty. Additionally to this externally-forced weakening, the AMOC exhibits internal variability which is independent of any changes in the boundary conditions. One pronounced mode of internal variability is the sub-decadal variability with a period of roughly 8 to 9 years. It is also apparent in the North Atlantic Oscillation (NAO) and in further parameters of the North Atlantic. The second part of this thesis describes the mechanism that produces this mode. Simulations from the Kiel Climate Model (KCM) indicate that the atmosphere-ocean interaction is crucial: Two feedbacks, a positive and a delayed negative one that involves the AMOC, act to generate the sub-decadal mode with the deep ocean setting the timescale. In the Southern Ocean, only one major event of open-ocean deep-convection occurred during the observational period. Still, tree-ring records from the southern hemisphere reveal a pronounced multi-centennial variability, which might be related to the recurring onset and shutdown of open-ocean deep-convection. This is also supported by models which simulate deep-convection variability on a wide range of timescales. The third part of this thesis investigates which key factors control this variability. Simulations from CMIP5 models and the KCM suggest that the mean state of the ocean stratification and of the sea ice volume are critical: Models with a strong (weak) mean stratification exhibit long (short) timescales of variability. Models with a large (small) mean sea ice volume tend to have the non-convective (convective) state as their default