Extended Data Fig. 9: Evolution in the sensitivity experiments.

a–g, Upper ocean (0–300 m) basin mean salinity budget for the subtropics (10° S–34° S) for ocean-alone sensitivity experiments. a, Control run annual mean budget. b,c,d, The accumulated anomalies (relative to control) for South Atlantic and South Indo-Pacific in experiments EmP+HFX (b), EmP (c) and HFX (d). Note the different vertical scales for each experiment. Over the South Atlantic, in HFX, the salinity divergence is reduced (positive in Extended Data Fig. 9d) by the weakening AMOC. With the combined forcing in EmP+HFX, the divergence of salinity transport still increases slightly over the South Atlantic (slightly negative in Extended Data Fig. 9b) as in the coupled model (Extended Data Fig. 7b), because the E−P forcing increases the salinity gradient and, in turn, the mean advection on the salinity anomaly and finally, the salinity divergence (Extended Data Fig. 9c). e,f,g, The time–latitude evolution of the AMOC (e) and upper (0–300 m) South Atlantic temperature (f) and salinity (g) in HFX shows a coherent penetration southward. The salinity response appears to respond earlier in the South Atlantic, likely caused by the divergence of the oceanic transport and salinity gradient. Therefore, the AMOC slowdown in the South Atlantic reduces salinity transport divergence, leading to the salinity pile-up there.