Picture (above): Underwater. Credit: Alexandra Koch (Pixabay).

In recent decades, oceanographers have observed localised events in polar regions where water near the surface of the ocean sinks rapidly. This phenomenon is known as open-ocean convection and it plays a key role in feeding a series of currents that together form a part of the global ocean ‘conveyor belt’. This conveyor belt, in turn, transports heat from the equator to the poles and keeps global temperatures relatively mild. Therefore, open-ocean convection plays a direct role in regulating our global climate.

However, a large proportion of global climate models do a relatively poor job of accurately representing the open-ocean convection we observe in the polar oceans. In the majority of models, the open-ocean convection is too strong, leading to excessively vigorous sinking of surface waters and flow-on effects on global ocean currents. In this paper, the researchers attempt to identify the cause of this excessive deepening and layout steps to address it in the future.

Ocean convection is inherently turbulent, that is, characterised by random and chaotic motions, often millimetres in size. Unfortunately, current computational resources, including supercomputers, do not allow researchers to model the ocean down to such a fine resolution, so they can’t accurately represent turbulence, and by extension, convection. Instead, they approximate the effect of turbulence in the ocean using parameterisations.

These parameterisations are not exact, so the researchers began their work by hypothesising that there may an issue in the convection parameterisation used in most ocean models, which leads to the anomalously strong sinking they produce. In order to test this hypothesis, they used a novel ocean model that does resolve turbulence down to its smallest size. They ran equivalent open ocean convection experiments with this model, known as a Direct Numerical Simulation (DNS), and a more ‘traditional’ ocean model, MOM6. Through this approach, they were able to identify the impact of accurately representing the turbulence in ocean convection.

’The evolution of the ocean’s mixed layer during ocean convection, as seen in a DNS model. Convection is partitioned into two distinct stages: deep convection, where the mixed layer grows vertically, and lateral spreading, where the mixed layer spreads horizontally.’

By comparing the two models, the researchers found that the current generation of convection parameterisations fail to replicate the random, chaotic nature of real-life turbulent convection. The lack of turbulence in MOM6 delays the formation of flow processes, which are important in halting the sinking of surface waters. Therefore, the surface waters sink too far compared to the real ocean and the DNS model.

In order to address the convection bias in ocean models, the researchers propose the development of a convection parameterisation that recreates the random nature of turbulent convection. Developing this new convection parameterisation will be undertaken in future research.