Abstract/Summary

The Atlantic Meridional Overturning Circulation (AMOC) is a key contributor in the climate system, transporting warm waters northwards throughout the Atlantic to compensate for the southward export of cold North Atlantic deep waters. However, observing the AMOC is inherently challenging due its long timescales and large spatial extent, which requires the use of time-evolving, three-dimensional model products to assess the ocean circulation, including: free-running models (FRMs), ocean reanalyses (ORAs) and coupled climate simulations. In the first and second chapters of this thesis, we show that the present ocean observation network and data assimilation schemes can be used to consistently constrain the ORA interior circulation in both structure and strength throughout the South Atlantic, which is improved with respect to the FRMs. This is in striking contrast with the large disagreement found in the ORA AMOCs, which is traced back to transport discrepancies in the South Atlantic western boundary currents at both upper and deep levels, explaining up to 85% of the ORA spread in the meridional heat transports. This will likely limit the effectiveness of ORAs for climate or decadal prediction studies. Using the same FRMs and ORAs with realistic salinities but quite different AMOCs, we show in the third chapter that the fresh Antarctic Intermediate Water layer eliminates salinity differences across the AMOC branches at �1200 m, which decouples the overturning freshwater transport Fov from the AMOC south of �10�N. In the South Atlantic the 0-300 m zonal salinity contrasts control the gyre freshwater transports Fgyre, which also determine the total freshwater transports. This decoupling makes the southern Fov unlikely to play any role in AMOC stability, leaving indirect Fgyre feedbacks or Fov in the north, as more relevant factors. In the fourth chapter of this thesis, the internal variability of 10 pre-industrial coupled simulations are used to show that coupled climate systems are governed by different dynamics and contain feedbacks not captured by the box-models. The spatiotemporal variability of the Atlantic freshwater budgets reveals that the in uence of Fov at 34�S in explaining freshwater content (FWC) changes is restricted to the South Atlantic and is always smaller than the in uence of Fgyre at 34�S. This consistently refutes the main salt-advection feedback mechanism, which states that Fov at 34�S is the main driver of meridional FWC perturbations. These results represent a substantial step in improving the understanding of the AMOC stability and have an important impact on climate modelling, since FRMs, ORAs and climate models refute box-model theories of AMOC bi-stability and suggest that other climate feedbacks associated with Fgyre and the atmosphere will likely dominate those associated with Fov throughout the South Atlantic.