This indicates that the change in atmospheric horizontal resolution also plays an important role, as explained in Hourdin et al. (2012) and also underlined by Marti et al. (2010). Note also that these numbers highlight the fact that CM5_piStart

(and thus also probably CM5_RETRO) is not in full equilibrium, as the intensity of the flow through the Drake Passage in this simulation (109 Sv) slightly differs from what is found in CM5-piCtrl (98 Sv). Finally, intensification of the ACC in CM5_piStart is consistent with strengthening of the density gradient across the Southern Ocean, as described above (Fig. 10), but this does not allow distinguishing causality. On the other hand, it contrasts with the weaker eastward heat transport seen in Fig. 11, demonstrating the importance of meridional temperature gradients and meanders for this heat transport (Sun and Watts, 2002). Downstream of the Drake Passage, the circumpolar ABT-263 concentration transport of mass is fed BIBW2992 concentration in both simulations by a weak input from the South Atlantic and a stronger one from the Indian Ocean, consistent with inversions from Ganachaud and Wunsch (2000). In both simulations, it thus slightly increases from the Drake Passage to the Cape of Good Hope and Cape Leewin sections. In the Pacific, the net mass transport is northward at all latitudes. This is again consistent with Ganachaud and Wunsch (2000). Their inversion yields an Indonesian Throughflow of 16 Sv, and the latest long-term

simultaneous measurements within both inflow and outflow passages (INSTANT

2004–2006) estimated a total transport of 14 ± 3 Sv (Sprintall et al., 2009). The intensity of the Indonesian Throughflow in terms of net mass transport in CM5_piStart is lower than these estimates (12.7 Sv Fig. 13), slightly stronger than in CM5_RETRO (12.3 Sv), although the difference is probably not significant. This might be again a consequence from the implementation of the ITF parameterisation developed by Koch-Larrouy et al. (2009) in CM5_piStart. Fig. 14 (left panels) compares the global mean meridional circulation for the years [2200–2291] of each simulation. The major difference lies in intensification by roughly 12 Sv of the Antarctic Bottom Water circulation in the Southern Hemisphere in CM5_piStart as compared to CM5_RETRO. This increase is roughly portioned Autophagy activator among the oceanic basins according to their width, as this cell increases by 4 Sv at the southern bottom of the Indian Ocean, 5 Sv in the Pacific and only 3 Sv in the Atlantic (not shown). As seen above for the forced simulations, this intensification is consistent with the evolution of the oceanic component, and in particular the implementation of the kz-tide parameterization. It is associated with notable temperature and more importantly salinity modifications in the Southern Ocean and along the sea floor, as described above. The shallow subtropical cells are very similar in the two simulations, consistent with comparable mean wind stress field and wind stress curl (not shown).