Sea levels of different atmosphere–ocean general circulation models (AOGCMs) respond to climate change forcing in different ways, representing a crucial uncertainty in climate change research. We isolate the role of the ocean dynamics in setting the spatial pattern of dynamic sea-level (ζ) change by forcing several AOGCMs with prescribed identical heat, momentum (wind) and freshwater flux perturbations. This method produces a ζ projection spread comparable in magnitude to the spread that results from greenhouse gas forcing, indicating that the differences in ocean model formulation are the cause, rather than diversity in surface flux change. The heat flux change drives most of the global pattern of ζ change, while the momentum and water flux changes cause locally confined features. North Atlantic heat uptake causes large temperature and salinity driven density changes, altering local ocean transport and ζ. The spread between AOGCMs here is caused largely by differences in their regional transport adjustment, which redistributes heat that was already in the ocean prior to perturbation. The geographic details of the ζ change in the North Atlantic are diverse across models, but the underlying dynamic change is similar. In contrast, the heat absorbed by the Southern Ocean does not strongly alter the vertically coherent circulation. The Arctic ζ change is dissimilar across models, owing to differences in passive heat uptake and circulation change. Only the Arctic is strongly affected by nonlinear interactions between the three air-sea flux changes, and these are model specific.

Four repeat hydrographic sections across the eastern Weddell gyre at 30°E reveal a warming (by ~0.1°C) and lightening (by ~0.02–0.03 kg m−3) of the Antarctic Bottom Water (AABW) entering the gyre from the Indian sector of the Southern Ocean between the mid‐1990s and late 2000s. Historical hydrographic and altimetric measurements in the region suggest that the most likely explanation for the change is increased entrainment of warmer mid‐depth Circumpolar Deep Water by cascading shelf water plumes close to Cape Darnley, where the Indian‐sourced AABW entering the Weddell gyre from the east is ventilated. This change in entrainment is associated with a concurrent southward shift of the Antarctic Circumpolar Current's (ACC) southern boundary in the region. This mechanism of AABW warming may affect wherever the ACC flows close to Antarctica.

<p>A rise in global mean sea level is a robust feature of projected anthropogenic climate change using state-of-the-art atmosphere-ocean general circulation models (AOGCMs). However, there is considerable disagreement over the more policy-relevant regional patterns of sea level rise. The Flux-Anomaly-Forced Model Intercomparison Project (FAFMIP) aims to improve our understanding of the mechanisms controlling regional and dynamic sea level change. In FAFMIP, identical air-sea buoyancy and momentum flux perturbations are applied to an ensemble of different AOGCMs, to sample the uncertainty associated with model structure and physical processes. Our novel implementation applies FAFMIP perturbations to an ensemble of OGCMs. This framework enables an estimate of the unknown atmosphere-ocean feedbacks, by comparing the coupled and ocean-only response to surface flux perturbations.</p><p>Comparing the response to idealised FAFMIP forcing with more realistic, increasing CO2 forcing, much of the spread in regional sea level projections for the North Atlantic and Southern Ocean arises from ocean model structural differences. Ocean-only simulations indicate that only a small proportion of this spread is due to differences in the atmosphere-ocean feedback. Novel tendency diagnostics indicate the relative effect of resolved advection, parametrised eddies, and dianeutral mixing on regional and dynamic sea level change. This study helps to reduce uncertainty in regional sea level projections by refining our estimates of atmosphere-ocean feedbacks and developing our understanding of the physical processes controlling sea level change.</p>