Fig. 3: SnSiZer analyses performed on proxy records from the West Mediterranean, West Africa and West-Central Africa during the last 70 kyr.

a West Mediterranean MAP stack (mm/yr). b–e West African records: b Stack of dust% (normalized, inverted axis). c Stack of ²³⁰Th-normalized dust flux (normalized, inverted axis). d Humidity Index record. e Stack of δD leaf waxes (normalized, inverted axis). f West-Central African MAP stack (mm/yr). The left panels display the raw data, while the right panels show the SnSiZer graphs based on the raw data. In the SnSiZer graphs, blue and red clouds indicate statistically significant decreasing and increasing rainfall, respectively. The vertical axis represents the level of smoothing in logarithmic units. Dark gray areas represent non-significant changes, whereas light gray areas denote low sampling resolution. In both the left and right panels, the pink-blue-yellow lines correspond to different Nadaraya-Watson smoothing levels for each record. Statistically significant changes in the data occur when a given smoothing level (pink, blue or yellow horizontal lines) intersects a statistically significant feature (red or blue clouds). These intersections are marked with blue and red arrows, indicating statistically relevant periods of decreasing and increasing rainfall, respectively. To identify a Heinrich Stadial as statistically significant, the analysis must show a statistically relevant decrease in rainfall followed by a significant increase a few thousand years later. Vertical yellow shades show the age-range of the Heinrich Stadials (HS6-HS1) and Younger Dryas (YD), as defined by various studies (see Supplementary Table 1 for age-ranges).