Environmental conditions associated with initial northern expansion of anatomically modern humans

The study addresses the long-standing debate over whether the initial dispersal of anatomically modern humans beyond Africa was primarily driven by environmental determinants or by cultural and social factors. Multiple dispersal events are likely, with early occurrences across southwest Asia, China, and Europe, and a consensus that major exits from Africa occurred between >100 ka and ~65 ka. Reconstructing the relative roles of ecological versus cultural processes has been hampered by sparse and discontinuous archaeological and genetic data, dating uncertainties, taphonomic biases, and potential lack of genetic contribution from earliest settlers. Existing modelling often embeds untested demographic assumptions and is validated against sparse datasets. To resolve these issues, the authors develop a statistical framework that jointly leverages radiocarbon-dated archaeological evidence and present-day human mitochondrial genetic differentiation to infer continuous spatiotemporal maps of first appearances and more-likely expansion routes across northern Eurasia and the Americas. They hypothesize that if environmental suitability guided initial expansions, then preferred routes should exhibit distinct environmental signatures: mid-range temperature and precipitation, proximity to major rivers or coasts (freshwater and resource access), low ruggedness, and travel along forest–grassland ecotones (higher biomass and mobility advantages).

Prior research proposes that orbital-scale climate shifts created corridors for human expansion out of Africa and that water availability, temperature, physical barriers (glaciers, mountains), and ecosystem type constrained dispersal. Archaeological-only or genetic-only reconstructions are limited due to sparse records, turnover erasing signals, and potential non-contribution of early settlers to modern gene pools. Earlier models operate at coarse scales with demographic assumptions and validation against incomplete data. Statistical approaches to map continuous appearances/disappearances have emerged but had limitations with radiocarbon data and spatial bias. Evidence for presence in Europe by ~46 ka and high latitudes by ~32 ka, and for delayed North American entry (~16 ka) with rapid spread to South America by ~14.6 ka, frame the debate, while discussions continue over northern versus southern (Arabian–South Asian) routes and the roles of coastal migration and environmental adaptation.

Datasets: The authors compiled 24,524 radiocarbon-dated archaeological specimens indicating human presence from multiple databases (INQUA Palaeolithic Europe, PIDBA, CARD) and curated literature for Japan and South America. They applied stringent quality assessment (rank ≥13/17) considering dating method, material, association, and protocol issues, excluding anomalous or unreliable dates and transitional Middle–Upper Palaeolithic layers with ambiguous attribution. This yielded 5,977 reliable ages, calibrated with IntCal13/ShCal13 in OxCal. Genetic data comprised 67,643 mtDNA control-region sequences from GenBank (last updated May 2017), filtered to 27,506 sequences with reliable georeferencing and Indigenous lineages. Haplogroups were assigned with HaploGrep2 (PhyloTree Build 17) across 96 geographic units; pairwise FST distances were computed from haplogroup frequencies. Spatiotemporal inference: They adapted a maximum-likelihood approach that corrects for the Signor–Lipps effect to estimate first-arrival times at each location, incorporating normally distributed dating errors and spatial weighting (Gaussian kernel) with bandwidth optimisation. A simulation-based spatial bias-correction estimated and removed local bias and variance to minimise integrated mean-squared error, yielding spatially corrected maps of initial arrival times. Pathway analysis: At 1° × 1° resolution, seven source–destination pairs were defined (Fertile Crescent to Beringia, Scandinavia, Japan, Portugal; Beringia to Central America; Central to Eastern Brazil; Central America to Chile). A modified least-cost framework computed trajectory costs using either local differences in inferred arrival time or mitochondrial FST between neighbouring cells. A simulated annealing algorithm introduced stochasticity to explore multiple plausible routes and avoid purely goal-directed least-cost assumptions. For each pair, trajectories were ranked by cost and cross-proxy agreement, selecting the main (more-likely) route(s). Environmental analysis: Six variables were extracted along routes at the locally inferred time of arrival: mean annual temperature and precipitation anomalies relative to 90 ka; dominant vegetation (forest vs grassland) from BIOME4 simulations forced by HadCM3; distance to major rivers; distance to contemporaneous coastlines (accounting for sea level over 120 ka); and terrain ruggedness (Riley TRI from ETOPO1). Global and regional randomisation tests (10,000 permutations) compared environmental values along more-likely versus other plausible routes, normalising by route length and regional area, to estimate the probability that observed differences arose by chance.

Expansion routes: Joint archaeological and genetic analyses support a northern spread from Southwest Asia via the Caucasus and around the Black Sea into Western Europe, reaching Scandinavia (−48.3 ka) and via Dalmatian/Mediterranean coasts (−44.1 ka). Eastward, a key corridor skirted the Caspian Sea, bifurcating north of the Hindu Kush/Pamir toward Mongolia (−47.1 ka) and south of the Karakoram/Himalaya (−45.8 ka), with further pathways toward Japan (−44.2 ka) and Beringia (−34.7 ka). A separate route led through Altai/Transbaikal to Beringia. Americas: Entry into North America occurred via the Pacific Northwest coast at ~16.2 ka, followed by rapid expansion to South America by 14.8 ± 1.2 ka. A second entry through the Laurentide–Cordilleran ice-free corridor opened by ~13.4–12.6 ka (via the Mackenzie River), preceding the Clovis period. South American dispersal followed (i) a clockwise route across Brazil to the east Brazilian and southeast Argentine coasts and (ii) an anticlockwise route across the Andes to the Peruvian coast and south along the Pacific, reaching Patagonia by 14.4 ± 0.2 ka. Migration rates: Estimated expansion rates exceed those for Sahul (0.71–0.92 km yr−1) and European Neolithic farmers (~1 km yr−1) but align with hunter-gatherer mobility patterns (daily travel 0.4–15 km, occasional long moves up to 60–80 km day−1). Some fastest rates in Asia and South America likely reflect data sparsity. Environmental correlates (global randomisation; Table 1): Three variables most strongly differentiated more-likely from less-likely routes: proportion of forest (veg; probability not due to chance = 1.000), mean annual temperature anomaly (T; 0.977), and mean annual precipitation anomaly (P; 0.971). Distance to rivers showed moderate support (0.902), while coastal proximity (0.457) and ruggedness (0.000) were not significant globally. Quantitatively, more-likely routes tended to be up to ~3.5°C warmer, up to ~0.5 mm day−1 wetter, and traversed more open habitats (up to ~40% less forest cover) than alternative routes, while often remaining near forest–grassland ecotones. Regional patterns: River proximity exerted local influence under harsh conditions (e.g., arid northern Europe along the Dnieper; Pr(R)rand < 10−3). Northern paths around Mongolia were wetter and associated with forest edges/new forests (Pr(R)rand ≤ 0.02). In drier southern routes across Russia toward Beringia, proximity to major rivers (Lena, Amur, Yellow) was advantageous. For Europe’s Atlantic approach via the Danube, river proximity was not significant (Pr(R)rand = 0.36), consistent with advantages of Mediterranean coastal movement. Standstill: Temperature-driven barriers explain an ~18.4-thousand-year migration hiatus in Beringia before initial North American entry at ~16.4–16.2 ka, consistent with the Beringian standstill/incubation model and genetic evidence of a bottleneck with expansion after ~16–13 ka.

Findings support that at broad spatial scales, environmental determinism structured early human expansion routes. Preferred pathways were systematically warmer, wetter, and aligned with forest–grassland ecotones, providing higher biodiversity, shelter, fuel, and enhanced mobility/visibility. Proximity to major rivers shaped local routing, particularly across arid or dense-forest regions, while ruggedness did not exert a strong global constraint, consistent with evidence for high-altitude and high-relief traversals (e.g., Tibetan Plateau ~40 ka). In Eurasia, corridors via the Caucasus and around the Caspian facilitated split routes eastward, with riverine and climatic factors differentiating northern and southern branches around Mongolia. In the Americas, climatic warming opened the ice-free corridor millennia after initial coastal entry, consistent with genetic bottleneck and expansion dynamics. Overall, integrating archaeological and genetic proxies with palaeoclimate and vegetation reconstructions reveals that environmental conditions largely determined emergent migration corridors, while acknowledging that cultural decisions and interspecies interactions likely influenced movements at finer spatial and temporal scales.

The study presents a quantitative, continuous spatial framework that integrates vetted radiocarbon archaeology, present-day mitochondrial genetic differentiation, and palaeoclimate–vegetation reconstructions to infer initial human expansion routes across northern Eurasia and the Americas and to test associated environmental drivers. More-likely routes were warmer, wetter, and tracked forest–grassland ecotones, with river proximity affecting local pathways; ruggedness and coastal proximity were not globally decisive. The framework overcomes common biases (e.g., Signor–Lipps, spatial binning, uneven sampling) and demonstrates that environmental determinism explains broad-scale palaeo-human movement patterns. Future work incorporating ancient DNA, additional chronometric methods (e.g., OSL, U–Th) to include southern Eurasian routes, and expanded archaeological coverage (especially in Asia and South America) will refine spatiotemporal inferences and clarify interactions between environmental, cultural, and demographic processes.

Analyses are constrained by the ~50 ka limit of radiocarbon dating, excluding many southern Eurasian regions primarily dated by OSL/U–Th. The focus on northern routes omits substantial areas (≈9.5% of the study’s potential land area) including much of Arabia, South Asia, and Southeast Asia. Sparse and uneven archaeological coverage in parts of Asia and South America may inflate local expansion-rate estimates. Genetic inferences rely on present-day mtDNA control-region haplogroup frequencies and FST, which may not fully capture ancient population structure or turnover; ancient DNA was excluded due to scarcity. The approach primarily detects sustained settlement rather than brief migration pulses. Assumptions in assigning Middle vs Upper Palaeolithic industries to Neanderthals vs Homo sapiens may not hold uniformly before 50 ka. Coastal proximity showed limited global signal, potentially due to sea-level change complexities and preservation biases along paleo-shorelines.