Climate change, not human population growth, correlates with Late Quaternary megafauna declines in North America | Nature Communications

Our results are at odds with simple overkill models that imply that multiple North American megafauna were directly driven to extinction by unsustainable hunting of rapidly expanding human populations6. Likewise, while recent studies have often emphasised that both overkill and climate change played a role in the extinction of different species of megafauna24, our analysis failed to replicate this finding and instead found a consistent correlation only between climate change and North American megafauna population levels. This was the case regardless of whether we analysed all megafauna together, or separated megafauna by taxa and region. It remained the case in our extended analysis involving the annually-resolved NGRIP climate proxy for which we also accounted for chronological and measurement uncertainty.

The divergence between earlier findings and our own is likely the result of problems with the use of SPDFs as a population proxy30,31,34. As discussed above, this approach dubiously conflates process variation—i.e., through-time changes in population level—with the chronological uncertainty inherent in radiocarbon dates, which has significant analytical consequences for studying population dynamics. As recent simulations studies have shown31,35, attempting to explain through-time population level fluctuations by comparing this proxy to some covariate (e.g., temperature, another SPDF, etc.) can produce misleading results. Our findings show that this extends to the study of Late Quaternary North American megafauna extinctions, and calls into question the use of SPDFs for studying extinction dynamics.

It is also important to recognise the problems with the radiocarbon record. While North America has some of the most detailed Late Quaternary archaeological and palaeontological records, samples sizes remain limited given the vast spans of space and time involved, and there are a number of important sources of bias (e.g., through-time taphonomic degradation of samples, spatio-temporal sample adequacy, radiocarbon-date calibration artefacts, etc.) that have implications for downstream analyses. Unsurprisingly, debates surrounding the chronology of human arrival to the Americas, founding population size, and subsequent population size fluctuations have continued with no sign of resolution. Concerning the megafauna record, it has long been known that the number of fossil finds in a given region/time to some extent corresponds with archaeological and geological research efforts aimed at dating material thought to be contemporaneous with humans in the Americas45. Although we attempted to correct for taphonomic and sampling biases in our analyses (see Methods section), there may still be biases that are difficult to control for at the moment and more research and data are needed.

Nevertheless, our findings make it clear that overkill by rapidly expanding human populations is not supported by the available data. Using the largest assembled database of directly dated North American megafauna, and accounting for chronological uncertainty in the radiocarbon and climate records, our results demonstrate that there is currently no evidence for a persistent through-time relationship between human and megafauna population levels in North America.

Our results are instead compatible with several alternative hypotheses. One is that while climate change appears to have been a dominant driving force behind megafauna population level fluctuations, humans may have been involved in more complex ways than simple overkill models suggest. Indeed, scholars have proposed a number of ways by which humans could have had a significant impact on megafauna populations that do not invoke widespread overhunting and significant population growth. Some have suggested that the depletion of keystone megaherbivores—those animals that have a disproportionately large influence over their environment—led to significant cascading effects on local flora and fauna46,47, as is known to occur in contemporary ecosystems48,49. So, even a few hunters on the landscape targeting only particular species might have led to population declines among numerous megafauna species without any long-term increase in hunting pressure from a growing human population. While this may have been the case for megafauna more broadly, our data indicate that at least some species of megafauna declines occurred prior to declines in keystone megaherbivores. Specifically, final declines in horse and saber-tooth cat population densities significantly pre-dated those of mammoths and mastodons. In fact, these population declines occurred at a time of increasing mammoth and mastodon numbers, which is particularly interesting in the case of the saber-tooth cat, which is often considered to have been a specialised hunter of these very large animals50. Others have proposed that increased competition between humans and carnivores forced carnivores to turn to and intensify predation on other, smaller animals51,52. Greater interspecific competition among carnivores for a smaller and less diverse food source would have driven population declines among not only herbivores but also carnivores, whereas humans may have been able to sustain (or even increase) population sizes despite dwindling megafauna numbers by exploiting a broad range of animal and plant foods50. Interestingly, there does appear to be a drop in saber-tooth cat population density coinciding with the emergence of Clovis-point wielding peoples in the Americas suggesting that interspecific competition may have had an initial impact on saber-tooth cat populations; although, the sample size for saber-tooth cat is rather small, and the final population decline appears to have occurred closer to the Younger Dryas (YD). Others still have suggested that humans, through hunting and habitat fragmentation, interrupted megafauna subpopulation connectivity, fragmenting populations into smaller, non-viable groups40,53,54. Indeed, megafauna, with their large home ranges, small population sizes, and slow life histories are particularly susceptible to extinction by habitat and population fragmentation46. If so, the mammoth and mastodon data suggest that this occurred not with the arrival of Clovis-point wielding people, but much later during the YD.

Alternatively, climate change may have indeed been the primary driving force behind the extinctions, with humans playing no significant role, or perhaps at most performing a coup de grâce on megafauna populations already heading towards extinction. Two key climatic events are often emphasised in the extinction of North American megafauna—the warm Bølling-Allerød interstadial (B-A; ~14.7–12.9 ka) and the cold YD stadial (~12.9–11.7 ka). Indeed, of the 37 genera that went extinct during the late Pleistocene, 16 have last appearance datums (LADs) that fall between 13.8–11.4 ka55, encapsulating the B-A/YD boundary.

Hypotheses focusing on the B-A assert that rapid temperature increase and associated ecological changes led to the extinction of North American megafauna. Some scholars have argued that the abrupt warming associated with interstadials drove megafauna extinctions across the Americas and Eurasia40,41. In North America, for instance, Cooper and colleagues40 posited that megafauna extinctions corresponded with or closely followed the abrupt warming of the B-A, and similarly timed megafauna population declines have been inferred from declines in Sporormiella spore abundance, a fungus found in the dung of ungulates and used as a proxy for megafauna population level changes56,57. Even though our analysis identified a positive relationship between temperature and megafauna population levels, it’s possible that a non-linear relationship exists. Megafauna populations levels could, for example, have generally increased along with temperature, giving rise to the relationship we identified, but then crashed in response to rapid temperature shocks (like the onset of the B-A) that crested some currently unidentified eco-biological threshold. Further, the extreme temperature upturns that characterise the Pleistocene interstadials are typically followed by more gradual temperature downturns. This means that a rapid increase may trigger ecosystem changes that precipitate population declines, which, in turn, correspond to subsequent temperature declines giving rise to the positive correlation between the climate and population proxies. Currently, the fossil evidence is not sufficient to evaluate this non-linear model, and our analysis cannot rule it out.

That said, we think our findings point more clearly to the onset of the YD as driving megafauna declines and extinctions. As can be seen in Fig. 5, the density of megafauna dated events increases during the B-A before decreasing shortly after the onset of the YD. The same can be said independently for mastodon, saber-tooth, and sloth populations, whereas final mammoth population declines appear to have occurred later in the YD, and final horse population declines appear to have occurred during the terminal B-A (Fig. 7).

Fig. 7: Climate, human and megafauna population changes through time.

Annual NGRIP oxygen isotope (δ18O) record with temporal uncertainty re-projected into the measurement domain (top) and Radiocarbon-dated Event Count Ensembles (RECEs) for humans, megafauna, and megafauna broken down by taxa (bottom). For more details see Fig. 5.

Full size image

While some scholars have dismissed the YD as a driver of megafauna extinctions, arguing that conditions were no more severe than earlier cold periods through which North American megafauna survived (e.g., the Last Glacial Maximum [LGM])1, there is evidence to suggest that the YD involved a specific set of climatic and ecological changes that may have been particularly devastating to megafauna populations58. During the YD, summer insolation (seasonality) reached one of its highest peaks of the late Pleistocene59, atmospheric CO2 concentrations rapidly rose60, and some of the fastest vegetation changes of the Late Glacial occurred61,62. Increased seasonality would have brought about shorter peak plant growth seasons, declines in plant community heterogeneity, and shifts in plant anti-herbivory defences63,64. Consistent with the fossil record, these conditions favoured smaller-bodied ruminants (e.g., moose, white-tailed deer) which can subsist on less diverse diets and are better equipped for exploiting plant nutrients across shorter growing seasons than larger-bodied monogastric animals and/or those with slow life histories (e.g., mammoth, mastodon, sloth)64. Increased atmospheric CO2, which is known to reduce plant nitrogen content, may have led to poorer quality forage and reduced landscape carrying capacities65. And in some regions, major vegetation changes in response to climate change were so rapid (<100 years)59,61,62 that megafauna populations may not have had time to adapt. Indeed, some studies have found declines in animal and plant biodiversity66,67 and genetic diversity37 occurring around the onset of the YD which have been tied to climate change and, occasionally, to more catastrophic events such as an extra-terrestrial impact68, although the latter are often disputed69.

The conditions of the YD were, however, not felt evenly across North America58, and, unsurprisingly, animal responses to these changes varied across both time and space e.g. refs. 24,37,70. On that note, it is important to consider the evidence divided into the Southwest and Great Lakes regions, drawing on the available local palaeoclimatological and palaeoecological records and our results.

In the Great Lakes region, mastodon and mammoth show an overall increase in dated event counts during the B-A and a decrease coinciding with the onset of the YD or shortly afterwards (Fig. 8). Pollen, stable isotope, and lake level data indicate cooler (~5 °C) and drier conditions in the Great Lakes region during the YD71. Vegetation changes were rapid, both in terms of their abruptness (within a century of the YD onset) and rate at which vegetation expanded geographically (>300 km/century)59. While a number of scholars have argued that megafauna declines significantly pre-dated major plant community changes in the Great Lakes region56,57,72, our findings suggest that YD climate, megafauna population, and plant community changes closely tracked one another. Open spruce (Picea) and sedge dominated parklands were quickly replaced with westward migrating mixed pine (Pinus) forests, with corresponding increases in taxa such as birch (Betula), elm (Ulmus), and oak (Quercus)56,57,59,61,72. The reduction of open environments in the case of mammoth, and of spruce woodlands in the case of mastodon, may explain the extirpation of these megafauna from the region73.

Fig. 8: Climate, mammoth population and mastodon population changes through time in the Great Lakes (GL) region.

Annual NGRIP oxygen isotope (δ18O) record with temporal uncertainty re-projected into the measurement domain (top) and Radiocarbon-dated Event Count Ensembles (RECEs) for mammoth and mastodon for the Great Lakes region (bottom). For more details see Fig. 5 caption.

Full size image

Likewise, in the American Southwest, sloth and mammoth show an overall increase in dated event counts during the B-A and a decrease around the onset of the YD (Fig. 9). That megafauna populations seemingly increased during the B-A is interesting given that speleothem and paleolake records indicate that the region experienced significant aridification74,75,76, being described by some as the worst drought in the American Southwest in 46,000 years77. Most speleothem and lacustrine records indicate cooler and wetter conditions following the onset of the YD e.g. refs. 74,75,77,78,79,, although it has been suggested that the Southern High Plains experienced some periods of drought80. Plant community responses to these climate changes appear to have been highly variable and/or poorly resolved and “consensus does not exist regarding the magnitude of direction of YD climate change in the Southwest” (Ballenger et al.81, p. 511). For instance, some plant pollen and macrofossil records indicate a transition from open environments to woodlands dominated by pine, oak, and juniper81,82, others indicate a somewhat opposite trend83, and others still show little change across the B-A/YD transition84. Attempting, then, to explain the American Southwest megafauna population declines as a response to changes in plant communities is, at present, difficult.

Fig. 9: Climate, mammoth population and sloth population changes through time in the American Southwest (SW).

Annual NGRIP oxygen isotope (δ18O) record with temporal uncertainty re-projected into the measurement domain (top) and Radiocarbon-dated Event Count Ensembles (RECEs) for mammoth and sloth for the Southwest region (bottom). For more details see Fig. 5 caption.

Full size image

In summary, the results of our quantitative analyses are consistent with climate-driven declines in North America’s megafauna populations. Data quality issues aside (see Introduction), using the largest assembled database of directly dated megafauna, we found no through-time relationship between megafauna and human population levels. While this does not preclude humans from having had an impact—for example, by interrupting megafauna subpopulation connectivity or performing a coup de grâce on already impoverished megafauna populations—it does suggest that growing populations of “big-game” hunters were not the primary driving force behind megafauna declines and extinctions. Instead, we found a consistent positive correlation between megafauna population levels and the NGRIP climate proxy. In other words, decreases in global temperature correlate with decreases in megafauna population levels. Final megafauna population declines leading to extinction roughly coincided with the onset of the YD, hinting that the unique conditions of the YD—i.e., abrupt cooling, increased seasonality, increased CO2, and major vegetation changes—played an important role in the extinction of North America’s megafauna. Furthermore, these findings suggest that YD climate, megafauna population, and plant community changes were in approximate equilibrium.

The causes(s) of North American Late Quaternary megafauna extinctions are likely to remain contentious. While to many researchers it may be an unlikely coincidence that megafauna extinctions coincided with human arrival at different times and in different parts of the world, it remains important to scientifically demonstrate this. And in doing so the limitations of the record are readily apparent: we simply do not have robust records for fauna and humans for vast spans of time and space. Building reliable records, and developing robust methods for interpreting them, remains a key task.