The genus Homo has existed for some three million years^4,5. For one third of this stretch of time, human species were confined to tropical and sub-tropical Africa, which is the homeland of the genus^6,7 and is rich in the warm, savanna-like environments to which most early hominins were best adapted^8,9. With the emergence of Homo erectussome 2 Ma, our ancestors began to disperse outside of Africa but kept away from boreal latitudes, possibly because of physiological limits to cold tolerance¹. However, later Homo species were able to expand their distribution to Northern Europe and Western Siberia, despite the advent of full glacial cycles which were making global temperatures colder than ever before during the history of the genus Homo . Findings in Happisburgh and Pakefield (UK) date the earliest presence of Homo at the southern edge of the boreal zone at some 0.7-0.9 Ma¹⁰. The occupation of such northern temperate and boreal zones presents a number of notable challenges. Not only does the cold itself present a challenge for hominins physiologically adapted to environments in Africa, but seasonality imposes extreme annual resource fluctuations, which imply a reliance on hunted meat for survival⁷ and limits to population densities which threaten the maintenance of mating networks. Technological adaptations facilitating survival in cold environments may have included the use of fire, shelters or clothing, with vulnerable infants being particularly susceptible to mortality. Technological advances such as the use of spears may have facilitated the hunting of typically large mammals¹¹ whilst social adaptations to care for higher injury rate may also have been important². At the same time cultural changes may have been necessary to enable knowledge transmission over large regions, and across generations, with archaeological evidence suggesting the emergence of larger networks of communication^1,2.

Clothing manufacturing leaves very little in the way of fossil remains¹². The first microwear evidence of hide-scraping (for manufacturing clothes) at Hoxne (UK), Biâche-Saint-Vaast, Pech de l’Azé and Abri Peyrony (France) and Shöningen (Germany)^13-15 are just some 50ka old at the most. Only the two most recent human species, H. neanderthalensisand H. sapiens , left incontrovertible evidence that they were able to produce complex, cold-proof clothing at that time. To make things more complex, in the particular case of H. neanderthalensis biological adaptation, besides material culture, was possibly involved in their ability to withstand the cold. Neanderthals possessed relatively short limbs, and a large midface and nasal cavity proposed to be specific cold adaptations, to heat and humidify inspired air, although the issue is definitely unresolved^16,17. In contrast to any other Homo , H. sapiens is considered the only species in the genus able to occupy cold regions through a genuinely cultural process, driven by our technology including the mastering of fire, ever-improving clothing craftsmanship and construction of shelters^15,18-20. This view setsH. sapiens apart from any other human species in terms of cognitive skills and implicitly rejects the idea that older Homomay have had sufficiently modern material culture to overcome climatic harshness²¹. With such a poor fossil record of clothes, tools to produce them and great uncertainty about deep-past local paleoclimates and human dispersal timing and direction, the issue of when humans first became cognitively and culturally able to extend their climatic tolerance beyond their physiological limits remains very difficult to decipher.

Here, we address the more restricted issue of when during the history ofHomo the limits of climatic tolerance expanded, and which species were involved. We do not specifically address the cultural and social adaptations that might underlie such tolerance, but rather consider the implications of our findings for the timing of such adaptations. We model climatic tolerance limits by associating palaeoclimatic values with the archaeological record to resolve the detail of the tempo and mode of past occupation of cold climates. Specifically, we test the hypothesis that H. sapiens developed greater climatic tolerance relative to H. heidelbergensis and H. neanderthalensisagainst the alternative that the exploration of climates outside natural physiological limits had already happened since the earliest of these species.

To test this, we estimated the rate of change of climatic tolerance limits across the human phylogenetic tree and searched for possible shifts in the rate, applying a method which allowed the rates to be estimated at each branch in the tree. In the present context, shifts in the rate of evolution of climatic tolerance that accrue to the clade including the Happisburgh/Pakefield hominins, H. heidelbergensis , plus H. neanderthalensis and H. sapiens , would indicate these hominins first acquired the capacity to develop cold-facing technological skills and cultural adaptations. Conversely, if either no rate-shift occurs, or the rate shift coincides with different clades (e.g. at the root of Homo ) the colonization of Northern habitats would be not indicative of any sudden increase in the ability to face environmental harshness.

The human fossil record dataset we used includes 2,597 hominin occurrences associated with 727 archaeological sites. The time range of our record spans from the first occurrence of australopiths in East Africa dated to some 4.2 Ma, to the definitive advent of H. sapiens in Eurasia almost coincident with the demise of H. neanderthalensis dated some to 0.040 Ma (see Supplementary Data). Such a wide range of hominin taxa provides a thorough phylogenetic context for the analyses.

Deriving spatio-temporally detailed climate data for the past requires dynamic climate modelling, but the timescales for human evolution exceed the possibilities of direct model simulation by several orders of magnitude. To circumvent this limitation, we combine direct simulation using a computationally efficient, intermediate complexity Earth system model, PLASM-GENIE, with statistical modelling, to create a PALEO-PGEM paleoclimate emulator, capable of performing multi-million year simulations forced by observationally derived proxy timeseries for ice-sheet state, CO₂ concentration and orbital forcing²². To model climatic niche evolution, we applied phylogenetic ridge regression (RRphylo )²³. RRphylo allows us to compute evolutionary rates for each branch of the phylogeny and to estimate the ancestral phenotypes. Here the ‘phenotype’ comprises climatic tolerance limits.

By using past seasonal maxima and minima for temperature, precipitation and annual net primary productivity from PALEO-PGEM, we reconstructed and projected onto the geographical space the climatic niche limits corresponding to the ancestral species distributions (the nodes in the tree) in our fossil database. Using RRphylo , we were then able to infer climatic niche tolerance limits for each node in the tree and to assess whether the rate of climatic niche evolution shows any shift (i.e. acceleration or deceleration) consistent with our starting hypothesis, while accounting for phylogenetic effects. We further accounted for phylogenetic uncertainty changing the tree node ages and the tree topology randomly one hundred times. By incorporating sources of uncertainty, we defined an overall ‘habitat quality’ metric, representing the number of iterations (out of 100) a geographic cell was found habitable (i.e. fell within climatic tolerance limits) for a given species (or ancestor in the tree).