Abstract. Climate variability on multidecadal timescales appears to be organized in pronounced patterns with clear expressions in sea surface temperature, such as the Atlantic Multidecadal Variability and the Pacific Decadal Oscillation.
These patterns are now well studied both in observations and global climate models and are important in the attribution of climate change.
Results from CMIP5 models have indicated large biases in these patterns with consequences for ocean heat storage variability and the global mean surface temperature.
In this paper, we use two multi-century Community Earth System Model simulations at coarse (1∘) and fine (0.1∘) ocean model horizontal grid spacing to study the effects of the representation of mesoscale ocean flows on major patterns of multidecadal variability.
We find that resolving mesoscale ocean flows both improves the characteristics of the modes of variability with respect to observations and increases the amplitude of the heat content variability in the individual ocean basins.
In the strongly eddying model, multidecadal variability increases compared to sub-decadal variability.
This shift of spectral power is seen in sea surface temperature indices, basin-scale surface heat fluxes, and the global mean surface temperature.
This implies that the current CMIP6 model generation, which predominantly does not resolve the ocean mesoscale, may systematically underestimate multidecadal variability.