Southern Ocean Centennial Oscillations in the CMIP6 Ensemble

Abstract

The Southern Ocean is a critical component in the global climate system, controlling key fluxes of CO2 and heat between the atmosphere and ocean interior, as well as the formation of water masses. In climate models, centennial-scale oscillations have been noted in key Southern Ocean properties such as the strength of the Antarctic Circumpolar Current, meridional overturning circulations, and deep ocean heat content, even in the presence of steady climate forcing. This thesis explores the dynamics behind these oscillations, how these vary across different climate models, and how imposed climate forcing may influence them. CMIP6 models were chosen as they are readily available and are a gold standard for climate projections used by the Intergovernmental Panel on Climate Change (IPCC) reports. The thesis begins with a thorough review of Southern Ocean centennial oscillations described in the literature, and proposed mechanisms including the "Heat Accumulation Release" hypothesis. This mechanism involves an increase in subsurface heat in the Southern Ocean due to incoming warm deep water, which gradually erodes the stratification until deep convection occurs. This deep convection ventilates the deep ocean, cooling it and resetting the stratification for the cycle to begin again. Throughout this thesis, measures of convection, stratification, temperature, current strength and water density are analysed to ascertain the dynamics and potential mechanisms of oscillation. The analysis begins with a pre-industrial control run of the CanESM5 model: a case study with clear centennial oscillations. Inconsistencies with previous literature hypothesis such as the heat accumulation release hypothesis are identified and a previously undescribed novel mechanism is described. This gyre advective mode is driven by the transfer of water masses between the subpolar gyres and the central ACC. This advection of water masses changes the stratification and convective behaviour of the southern Weddell, which feeds back onto the water mass properties, reinforcing the oscillation. The timescale appears to be set by the advective timescales of dense water exiting the gyre regions. This analysis is expanded to a selection of 16 CMIP6 models. Centennial oscillations in the Southern Ocean can be observed in twelve of these models, of which five (including CanESM5) are found to exhibit dynamics aligning with the newly described gyre advection mode. In contrast, three models oscillate via the previously described heat-accumulation release mechanism. Some models (HadGEM3-GC3.1-LL and IPSL-CM6A-LR) were found to transition between oscillatory behaviours. In all cases, oscillations were largely the result of Southern Ocean internal processes, associated with the movement of heat and vertical stratification, rather than external drivers such as atmospheric modes. The behaviour of convection could be used to separate models into oscillatory and non-oscillatory states, with models with strong, continuous convection not showing oscillations, and models with weaker convection being much more likely to show oscillations. However, no clear relationships were found between the state and configuration of the models and the oscillatory mechanism or strength. The behaviour of these oscillations under CO2 surface forcing scenarios is investigated in a subset of oscillating models, alongside novel runs of the oscillating HadGEM3-GC3-LL model which were produced to supplement this analysis with controlled initialisations from different phases of the oscillation. Surface forcing was found to be critical to the behaviour of the oscillations, with a one percent compounding increase in CO2 forcing able to suppress oscillations within 50 years by stratifying the surface and reducing convection. A halving of atmospheric CO2 was found to both increase the strength but also increase the frequency of oscillations. This analysis was then extended to the historical and future Shared Socio-Economic Pathway (SSP) scenarios for the CMIP6 ensemble. This showed that even weak future forcing (such as ssp119 and ssp126) is able to shut down deep convection within 30-50 years and thus stop the oscillations. This demonstrates the presence of a "tipping point" in the deep convection and lower circulation cell under future warming. The loss of oscillations means that the variation between runs decreases and they tend to converge within 60 years of the convection ceasing (approximately the time for the anomalous dense water in some models to advect out and the Southern Ocean states to converge). This means that the initial conditions and phase of oscillation in the scenario runs have little impact on the final state of the Southern Ocean, and variables such as the ACC. However, they dominate the uncertainty in the change in state from present day, often explaining >70% of this uncertainty in features such as the ACC transport.