Fig. 4: Paired temperature and hydroclimate reconstruction for easternmost Africa spanning approximately the past 75 kyr in relation to globally distributed palaeoclimate reference records.

a, Benthic marine oxygen-isotope stack²⁸, reflecting global ice volume. b, Synthetic Greenland oxygen-isotope record²⁶, reflecting Arctic temperature variation. c, Speleothem composite oxygen-isotope record from Hulu–Dongge⁵³, reflecting East Asian summer monsoon intensity. d, BIT index²⁰ of sedimentary GDGTs from Lake Chala, reflecting lake water-balance variation. e, MST variation based on sedimentary brGDGTs from Lake Chala, calibrated¹⁹ and rescaled (left y axis) using an eastern African ensemble reconstruction (Extended Data Fig. 4 and Methods); the purple shading shows the RMSE uncertainty envelope of this calibration, ±0.7 °C after rescaling. Data from sediment horizons partly consisting of thin turbidites are shown as grey crosses and excluded from interpretation. f, Atmospheric CO₂ concentration as recorded in Antarctic ice⁴⁴, with a shift from dark to light green centred on 250 ppmv. The ages of individual radiocarbon-dated intervals (thin vertical lines) and the 73.7-kyr YTT cryptotephra (grey diamond) used for age-model construction (Fig. 2 and Methods) are plotted along the bottom axis. The timing of the LGM, HS1 and YD are shown for reference. Considering linear age interpolation in the 25.2–73.7-kyr section of our proxy time series (Methods), the timing of Heinrich events H2–H6 should be viewed as approximative (as indicated by gradated light grey shading).