Extended Data Figure 9: Comparison of the AM variability with sea level (global ice volume) and atmospheric CO₂ changes.

Upper panel, interval from 350 to 0 kyr bp; lower panel, interval from 650 to 300 kyr bp. In both panels: a, AM Δδ¹⁸O record. b, AM δ¹⁸O record (green) and 21 July insolation at 65° N (pink)⁴⁵. c, Composite sea level record¹⁷. d, Composite atmospheric CO₂ record²⁰. Grey bars show the timing of WMIs and associated terminations. Two yellow bars indicate the two millennial-scale positive anomalies (or WMIs), marking the ‘unfinished terminations’³⁰—the MIS 4/3 and 5.2/5.1 transitions (Fig. 2). For the T-IIIa WMI we also indicate the correlation previously made by Cheng et al.⁴ with a beige bar. Although we consider this as a plausible alternative correlation, we prefer the new correlation presented in Figs 1, 3 and 5, and in this figure. The new correlation fits much better with the original chronologies of the ice core and marine records. In addition, the match of the adjacent high δ¹⁸O AM anomalies and the ice rafted debris record⁷ is better with the new correlation. Some examples of initial AM rises around NHSI minima are depicted by green arrows and dashed lines, which do not appear to link directly to either global ice volume or CO₂ changes.