This means that we may be able to say more about the future of the Southern Ocean than we might expect from "initial condition anomalies".

Therefore, when we look at the change from "present day" (the initial value anomaly), the high initial condition uncertainty is spuriously moved to our projections, when we actually know our current conditions well, and the projections agree on our future trajectories.

This means that variance in the raw data is much higher at the start than the end of the run, due to initial condition impacts.

This is key for all modellers projecting the Southern Ocean (and potentially more widely). As shown in the attached figure, the Antarctic Circumpolar Current shows great variability in the present day but much lower in the future as long-timescale variability shuts down.

I am very excited about this point: it makes a huge difference whether we analyse raw data, changes from the "present day", or from "pre-industrial" in our models. Changes from the "present day" can mean projecting forward current variability and add potentially spurious uncertainty to projections.

▪️ Consequences for Climate Projections: Long timescale variability can contribute the majority of uncertainty in climate projections.

▪️ Shutdown of the lower circulation cell: Under even weak future forcing the oscillations are damped out and the strength of the lower circulation cell decreases

The key points were:
▪️ Centennial Oscillations: Southern Ocean oscillations on hundred-year timescales exist in the CMIP6 ensemble of global climate models

Finally, the thesis also shows that under global cooling (perhaps due to a future overshoot in temperatures), the Southern Ocean becomes destabilised, triggering new and faster oscillations (such as in the Antarctic Circumpolar Current), and even a counterintuitive loss of sea ice in some sectors.

^^For more on this point, see our recent preprint: essopenarchive.org/users/946271...

As centennial variability dominates the uncertainty in some of the Southern Ocean, this shutdown in variability dominates the behaviour of uncertainty in features (e.g. the Antarctic Circumpolar Current), so the initial oscillation state accounts for a large amount of uncertainty in trajectories.

Applying global warming forcing to the system leads to the shutdown of convection and the loss of the oscillatory behaviour (shown in figure). Under plausible future scenarios (SSP scenarios for CMIP), this shutdown occurs even with strong mitigation, suggesting a tipping point may have been passed.

These oscillations involve multiple mechanisms, both the more traditional heat release behaviour and a novel advective mechanism involving the gyres. While both mechanisms involve deep convection, the advective mechanism involves motion of heat around the gyre.

Centennial (roughly hundred year period) oscillations exist in the Southern Ocean across the CMIP6 ensemble of climate models. These oscillations can be seen in many features such as the ocean heat, currents (such as the Antarctic Circumpolar Current, shown in the figure), and dense water formation.

Thesis link: www.repository.cam.ac.uk/items/f8e627...
I will try to outline some of the key takeaways from my thesis in the rest of this post:

Many thanks to my co-authors Andrew Meijers, Peter Haynes, and Mark Webb.
Additional thanks to the institutions that have funded and/or supported my PhD and this research: @granthamlse.bsky.social, @climtip.bsky.social, C-CLEAR DTP, @bas.ac.uk, @metoffice.gov.uk .

3) The loss of this variability dominates our ability to make future projections about features in the Southern Ocean, such as the Antarctic Circumpolar Current and ocean heat uptake.

2) We also demonstrate that these oscillations and the associated dense water formation cease under all future climate scenarios.

The three key takeaways are as follows:
1) We describe long-term convective oscillations seen in the Southern Ocean of the majority of the CMIP6 (global climate) models that we analysed.