C. Heuzé, H. Zanowski, S. Karam, and M. Muilwijk (2023), The deep Arctic Ocean and Fram Strait in CMIP6 models. Journal of Climate, vol 36, pp 2551–2584, doi:10.1175/JCLI-D-22-0194.1.
This paper describes biases in the CMIP6 models’ representation of the Arctic Ocean, from the Atlantic layer down. Although this study reports on only 14 models, we originally computed the biases on the same 36 models as Heuzé (2021).
- hydrographic biases: The Atlantic Water is too deep (400 m on average) and too cold (0.5 deg C on average), while the deep and bottom waters are too warm (1 deg C on average). Or rather, for most models, there’s no real distinction between the various water masses. The properties are also not changing throughout the Arctic, even though they should:
- ventilation: Two models clearly have dense water overflows in St Anna Trough, and another one might but it could not be detected with the monthly output. We found a strong disconnect between shelf properties / dense water overflows and polynya activity that needs further investigating. Only one other model has some relatively-deep convection instead of overflows.
- circulation within the Arctic: Not very well represented, but until we have better velocity data in the Arctic and more straightforward ways to un-rotate the model grids, we cannot be more specific. The age of water diagnostic was not useful as it was available for only half the models, and these models used different protocols anyway.
- Fram Strait: Volumes fluxes in and out are too weak, but Fram Strait itself is biased warm, so the heat fluxes appeared correct enough, albeit not necessarily at the correct depth or east-west location.
And now what?
For starters, we need more observations. Obviously the models are incorrect in the hardly-observed deep Arctic!
We also need more model diversity. Isopycnal and terrain following grids seemed better (as expected) at representing overflows. Higher resolutions, directly or via grid refinment, are required to better represent the bathymetry, which drives most of the circulation. And “polar-friendly” parameterisations need to be routinely deployed.
See also our companion paper Muilwijk et al. (2023) that details why these biases matter, focussing on future stratification and sea ice.