A comparison of historical North Atlantic variability in multi-cycle OMIP2 simulations at two different horizontal resolutions

Abstract:

The sensitivity to resolution of simulated historical AMOC and ocean northward heat transport (NHT) is examined using a pair of CMIP6-OMIP2 simulations performed with the CESM2 model: one at coarse resolution (nominal 1°; LR) and the other at eddy-resolving resolution (nominal 0.1°; HR). Both simulations are initialized from observed climatology and spun up through consecutive forcing cycles (6 cycles for LR; 4 cycles for HR) using JRA55-do atmospheric state fields and fluxes (1958-2018). The long spin-up allows for an assessment of how resolution impacts model drift and the signal-to-noise of surface-forced variability in the North Atlantic. Direct comparison to observed time series (including overturning measurements from RAPID and OSNAP; satellite-based sea surface height; and gridded temperature products) reveals the extent to which explicit simulation of ocean mesoscale turbulence improves overall simulation fidelity. While there is clear improvement in the representation of Labrador Sea processes in HR, bringing that simulation closer in line with OSNAP observations, AMOC/NHT variations at lower latitudes are not significantly different between HR and LR. In particular, the coupled overturning/gyre mechanism believed to underpin high decadal predictability in the subpolar Atlantic appears robust to changes in resolution.