The Neolithic transition has led to marked increases in census population sizes across the world, as recorded by a rich archaeological record. However, previous attempts to detect such changes using genetic markers, especially mitochondrial DNA (mtDNA), have mostly been unsuccessful. We use complete mtDNA genomes from over 1700 individuals, from the 1000 Genomes Project Phase 3, to explore changes in populations sizes in five populations for each of four major geographical regions, using a sophisticated coalescent-based Bayesian method (extended Bayesian skyline plots) and mutation rates calibrated with ancient DNA. Despite the power and sophistication of our analysis, we fail to find size changes that correspond to the Neolithic transitions of the study populations. However, we do detect a number of size changes, which tend to be replicated in most populations within each region. These changes are mostly much older than the Neolithic transition and could reflect either population expansion or changes in population structure. Given the amount of migration and population mixing that occurred after these ancient signals were generated, we caution that modern populations will often carry ghost signals of demographic events that occurred far away from their current location.

During the Quaternary, large climate oscillations impacted the distribution and demography of species globally. Two approaches have played a major role in reconstructing changes through time: Bayesian Skyline Plots (BSPs), which reconstruct population fluctuations based on genetic data, and Species Distribution Models (SDMs), which allow us to back‐cast the range occupied by a species based on its climatic preferences. In this paper, we contrast these two approaches by applying them to a large data set of 102 Holarctic bird species, for which both mitochondrial DNA sequences and distribution maps are available, to reconstruct their dynamics since the Last Glacial Maximum (LGM). Most species experienced an increase in effective population size (Ne, as estimated by BSPs) as well as an increase in geographical range (as reconstructed by SDMs) since the LGM; however, we found no correlation between the magnitude of changes in Ne and range size. The only clear signal we could detect was a later and greater increase in Ne for wetland birds compared to species that live in other habitats, a probable consequence of a delayed and more extensive increase in the extent of this habitat type after the LGM. The lack of correlation between SDM and BSP reconstructions could not be reconciled even when range shifts were considered. We suggest that this pattern might be linked to changes in population densities, which can be independent of range changes, and caution that interpreting either SDMs or BSPs independently is problematic and potentially misleading.

Extensive sequencing of modern and ancient human genomes has revealed that contemporary populations can be explained as the result of recent mixing of a few distinct ancestral genetic lineages1. But the small number of aDNA samples that predate the Last Glacial Maximum means that the origins of these lineages are not well understood. Here, we circumvent the limited sampling by modelling explicitly the effect of climatic changes and terrain on population demography and migrations through time and space, and show that these factors are sufficient to explain the divergence among ancestral lineages. Our reconstructions show that the sharp separation between African and Eurasian lineages is a consequence of only a few limited periods of connectivity through the arid Arabian peninsula, which acted as the gate out of the Arican continent. The subsequent spread across Eurasia was then mostly shaped by mountain ranges, and to a lesser extent deserts, leading to the split of European and Asians, and the further diversification of these two groups. A high tolerance to cold climates allowed the persistence at high latitudes even during the Last Glacial Maximum, maintaining a pocket in Beringia that led to the later, rapid colonisation of the Americas. The advent of food production was associated with an increase in movement2, but mountains and climate have been shown to still play a major role even in this latter period3,4, affecting the mixing of the ancestral lineages that we have shown to be shaped by those two factors in the first place.

During the glacial periods of the Pleistocene, swathes of the Northern Hemisphere were covered by ice sheets, tundra and permafrost leaving large areas uninhabitable for temperate and boreal species. The glacial refugia paradigm proposes that, during glaciations, species living in the Northern Hemisphere were forced southwards, forming isolated, insular populations that persisted in disjunct regions known as refugia. According to this hypothesis, as ice sheets retreated, species recolonised the continent from these glacial refugia, and the mixing of these lineages is responsible for modern patterns of genetic diversity. However, an alternative hypothesis is that complex genetic patterns could also arise simply from heterogenous post-glacial expansion dynamics, without separate refugia. Both mitochondrial and genomic data from the North American Yellow warbler (Setophaga petechia) shows the presence of an eastern and western clade, a pattern often ascribed to the presence of two refugia. Using a climate-informed spatial genetic modelling (CISGeM) framework, we were able to reconstruct past population sizes, range expansions, and likely recolonisation dynamics of this species, generating spatially and temporally explicit demographic reconstructions. The model captures the empirical genetic structure despite including only a single, large glacial refugium. The contemporary population structure observed in the data was generated during the expansion dynamics after the glaciation and is due to unbalanced rates of northward advance to the east and west linked to the melting of the icesheets. Thus, modern population structure in this species is consistent with expansion dynamics, and refugial isolation is not required to explain it, highlighting the importance of explicitly testing drivers of geographic structure.

No abstract