Extended Data Fig. 8: Climate effects on time-mean habitat suitability of Homo sapiens, Homo neanderthalensis, Homo heidelbergensis.

a–c) difference between time-averaged habitat suitability estimated from original climate envelope model H and climate envelope model H_scr calculated from vector-randomized climate variables (Methods) for 3 hominin groups. This type of time randomization (Methods) maintains the time-mean and climate co-variance among the 4-dimensional climate input vector. Results show the effect of the original orbital-scale climate trajectory (relative to a randomized climate trajectory) on hominin habitat suitability. Only grid points with p < 0.05 are shown in colors based on a paired t-test using H(x,y,t) and H_scr(x,y,t). d–f) Illustration of late-Holocene altitude-corrected downscaling of net primary production: Left: Simulated NPP (gC/m²/year) for original T31 atmosphere resolution, obtained from empirical NPP model using 1000-year late Holocene average of total precipitation and surface temperatures simulated by the CESM1.2 2Ma experiment; middle: same as left, but using altitude-corrections for temperature and precipitation as downscaling onto a 1x1 degree target grid, showing the emergence of key topographic features in Africa. This resolution was chosen in the calculations of the climate envelope model; right: for illustration, same as middle but for a 0.25x0.25 degree horizontal target grid. The qualitative gain in terms of regional details from T31 to 1x1 degree resolution outweighs the additional gain going from 1x1 degree resolution to a 0.25x0.25 degree grid.