On the implications of a warm bias in modelling an eddying Southern Ocean

Mesoscale eddies are considered to have a major impact on the horizontal and vertical redistribution of heat, freshwater, carbon and other passive tracers across the Southern Ocean (SO). A way to investigate the role of mesoscale dynamics in a region where observations are sparse is running a high-resolution model. Here, we apply 2-way nesting to the ocean model NEMO3.6 using its AGRIF module to simulate an eddying SO embedded in the fully coupled climate model FOCI. The nest enhances the horizontal ocean grid resolution from 1/2˚ to 1/10˚ everywhere south of 28˚S. Since the nested model, called FOCI-ORION10X, is computationally relatively expensive, our goal was to gain a spun up climate state with just the non-eddying resolution model (without nest). This would open the opportunity to efficiently run a coarse climate model into different climate states under which the role of mesoscale eddies could then be studied with the nested setup. Here, we demonstrate that there are limits to such an approach arising from the mean state of the climate model.

The non-eddying standard FOCI model features a significant warm bias in the SO similar to many CMIP-class climate models. To test the implications of the warm bias on the nested model configuration, we compare two such simulations branching off from coarse FOCI pre-industrial control simulations and contrast these to a nested run starting from rest initialized with Levitus (WOA13) temperature and salinity fields. The two FOCI control runs differ in warm bias intensity due to a shorter coupling frequency with the atmosphere and modified ocean mixing parameters. Further, one nested run is started already after 500 years from the weakly biased run and the other after 1500 years of the strongly bias run yielding a difference of ~50% in the temperature bias. In both cases, Weddell Gyre stratification becomes unstable within the first decade of the nested runs initiating open ocean deep convection and releasing the excess heat to the atmosphere. While the spurious deep convection results in a widely reduced heat bias in the nested runs after a few decades, it directly increases the meridional density gradient to the mid latitudes and enhances the strength of the Antarctic Circumpolar Current. Besides these positive effects, we also find unusually strong production of bottom water yielding a too strong bottom cell in the meridional overturning circulation. Especially because of this lasting deep ocean impact, we see no advantage in branching off from a biased mean state compared to the nested run starting from rest, which reaches a quasi-equilibrium after 100 years. We conclude, the presence of a typical warm bias and the SO’s sensitivity to stratification hinder the combination of eddying and eddy-parameterized model configurations to facilitate cost-efficient long spinup procedures.