Fig. 2: Reconstructed changes in ACC strength and compared with Antarctic temperature and Southern Ocean sea-ice extent.

a Antarctic temperature changes from the European Project for Ice Coring in Antarctica (EPICA) Dome C ice core47. b High-resolution XRF scanner-derived records of ln(Ca/Ti) (peak area count ratios; black triangles) were applied to fine-tune the Antarctic temperature anomalies together with radiocarbon dates (red triangles), relative paleointensity and paleomagnetic excursions (purple diamonds) age control points from core PS97/085-3 (see Supplementary Methods, Supplementary Figs. 2 and 3). c XRF-derived ln(Zr/Rb) variations indicate changes in sediment grain-size fractions. d Mean sortable silt grain size (\(\overline{{{{\rm{SS}}}}}\), 10–63 μm) reaches up its upper limit under high flow speeds. e Mean grain size of sortable silt and fine sand (\(\overline{{{{\rm{SSFS}}}}}\), 10–125 μm) was used as the ACC flow speed proxy in this study. f ACC bottom flow speeds were estimated by the correlation between the \(\overline{{{{\rm{SSFS}}}}}\) and adjacent current meter data (see Supplementary Fig. 6). g Sea salt sodium (ssNa+) flux from the EPICA Dronning Maud Land (EDML) ice core, a proxy for sea ice extent and atmosphere changes, smoothed with a three-points running mean34. Vertical gray bars mark inferred glacial periods and pink bars indicate the sub-interglacial during Marine Isotope Stage (MIS) 5.