From the Cover: Late Pleistocene shrub expansion preceded megafauna turnover and extinctions in eastern Beringia

The collapse of the steppe-tundra biome (mammoth steppe) at the end of the Pleistocene is used as an important example of top-down ecosystem cascades, where human hunting of keystone species led to profound changes in vegetation across high latitudes in the Northern Hemisphere. Alternatively, it is argued that this biome transformation occurred through a bottom-up process, where climate-driven expansion of shrub tundra (Betula, Salix spp.) replaced the steppe-tundra vegetation that grazing megafauna taxa relied on. In eastern Beringia, these differing hypotheses remain largely untested, in part because the precise timing and spatial pattern of Late Pleistocene shrub expansion remains poorly resolved. This uncertainty is caused by chronological ambiguity in many lake sediment records, which typically rely on radiocarbon (¹⁴C) dates from bulk sediment or aquatic macrofossils—materials that are known to overestimate the age of sediment layers. Here, we reexamine Late Pleistocene pollen records for which ¹⁴C dating of terrestrial macrofossils is available and augment these data with ¹⁴C dates from arctic ground-squirrel middens and plant macrofossils. Comparing these paleovegetation data with a database of published ¹⁴C dates from megafauna remains, we find the postglacial expansion of shrub tundra preceded the regional extinctions of horse (Equus spp.) and mammoth (Mammuthus primigenius) and began during a period when the frequency of ¹⁴C dates indicates large grazers were abundant. These results are not consistent with a model of top-down ecosystem cascades and support the hypothesis that climate-driven habitat loss preceded and contributed to turnover in mammal communities.

In northern high latitudes, the widespread extinction of Quaternary megafauna (animals weighing >44 kg) and disappearance of the steppe-tundra biome they inhabited is used as an important example of top-down ecosystem cascades, where human hunting of keystone species led to profound changes in vegetation structure at the end of the Pleistocene (15 thousand years before 1950 [15 ka] to 11.7 ka) (1–5). This hypothesis, however, is not well tested, and it is unclear whether the relative timing of megafauna extinctions and vegetation change is consistent with a top-down model. Resolving this question is an important part of understanding how past ecosystems functioned and may help predict how modern high-latitude ecosystems will respond to climate-driven vegetation change, current declines in large mammal species, or their deliberate reintroduction.In eastern Beringia (modern-day Alaska and the Yukon interior) (Fig. 1), Late Pleistocene megafauna extinctions broadly coincided with an expansion of shrub tundra vegetation including dwarf and tall-shrub species of birch (Betula nana and Betula glandulosa) and willow (Salix spp.) (6). Prior to these events, herds of grazing megafauna occupied a biome termed the mammoth steppe (7–9) or steppe-tundra (10, 11), which has no widespread modern analog. This novel, dry environment supported diverse plant communities, dominated by grasses, sedges, Artemisia spp., and a range of other forbs (8, 12–16). Sometime between 16 ka and 13 ka, woody shrub species began to expand across eastern Beringia, coupled with the development of peatlands and organic soil horizons (14, 17, 18). Lake sediment records show that the expansion of shrubs was rapid in many cases, and, although pollen influx data suggest herbaceous plant taxa continued to form an important part of the vegetation community, the abundance of Betula and Salix pollen (often >50% of the pollen sum) indicates that eastern Beringia became increasingly dominated by woody vegetation during this period (19–22).Open in a separate windowFig. 1.Eastern Beringia during the Late Pleistocene. Ice limits (14.2 and 15.5 ka) are redrawn from Dalton et al. (80). Lake sediment records reanalyzed in this study are numbered and include the following: 1, Burial Lake (81); 2, Tukuto Lake (82); 3, Lake of the Pleistocene (18); 4, Okpilak Lake (44); 5, Trout Lake (55); 6, Hanging Lake (54); 7, Ruppert Lake (83); 8, Xindi Lake (84); 9, Harding Lake (45); 10, Birch Lake (19); 11, Lost Lake (21); 12, Jan Lake (47); 13, Idavain Lake (20); 14, Beaver Lake (46); and 15, Discovery Pond (85).During the same time period, mammal communities in eastern Beringia underwent some of the most profound changes to occur in the region since at least the end of the last interglacial (Marine Isotope Stage 5e), 115 ka. Of the 13 megafauna taxa present in eastern Beringia immediately prior to 15 ka, only seven survived in situ beyond the Pleistocene (steppe bison, Bison priscus; caribou, Rangifer tarandus; wapiti, Cervus canadensis; muskox, Ovibos moschatus; wolf, Canis lupus; grizzly bear, Ursus arctos; and sheep, Ovis). The remaining taxa (caballine/stout-legged horses, Equus; stilt-legged horses, Haringtonhippus; woolly mammoth, Mammuthus primigenius; saiga antelope, Saiga tatarica; lion, Panthera spelaea; and short-faced bear, Arctodus simus), along with smaller mammals such as the arctic ground squirrel (Urocitellus parryii), became regionally extinct throughout large areas between 15.0 ka and 11.7 ka, leaving behind a comparatively impoverished mammal community (6, 23). The arrival of moose (Alces alces), an obligate browser, in eastern Beringia shortly after 15 ka (24) marks the beginning of a shift from the grazer community of the steppe-tundra toward a community of mixed-feeding megafauna species better adapted to a shrub tundra environment.The broad chronological overlap between the timing of shrub expansion and turnover in mammal populations has led numerous authors to hypothesize that habitat loss was a key driver of Late Pleistocene extinctions in eastern Beringia (8, 25–27). These authors argue that the Betula- and Salix-dominated shrub tundra was inhospitable to grazing megafauna because low-growing shrubs develop strong antiherbivory compounds, making them inedible or toxic to many mammals that lack a rumen to aid digestion (28). Other researchers have suggested that the decline in populations of grazing megafauna preceded shrub expansion, and that the spread of shrub tundra was caused by the resulting reduction in browsing pressure, vegetation trampling, and snow clearance (2, 29). These studies argue that grazers, and particularly megaherbivores (mammals of >1,000 kg), such as mammoth, acted as keystone species and were essential to the continuation of the steppe-tundra (1). In this case, human-caused “overkill” (30) or the compounded impacts of humans (e.g., burning, hunting, or simply their presence) in a dynamic ecosystem are advanced as the causes of megafauna extinctions. Finally, it is also possible that both of these processes reinforced one another, and the disappearance of the steppe-tundra was caused by both bottom-up and top-down pressures, or even that there was no causal relationship between the megafauna declines and shrub expansion. All of these hypotheses remain largely untested and continue to be controversial, in part because human arrival patterns, hunting preferences, and population size are largely unknown (31).In eastern Beringia, it is difficult to distinguish between these alternative hypotheses because the regional timing and spatial pattern of Late Pleistocene shrub expansion is poorly resolved, despite more than 50 y of detailed paleoecological study (14, 32–35). This uncertainty is principally due to the difficulty in accurately dating lake sediments from high latitudes (36–38). Terrestrial plant macrofossil remains are often rare in these depositional environments, and many pioneering paleoenvironmental studies are founded on chronologies based on radiocarbon (¹⁴C) dates derived from bulk sediment or aquatic macrofossils. This is particularly common for lake records obtained before the routine availability of accelerator mass spectrometer radiocarbon (AMS ¹⁴C) dating, when larger samples were required. Radiocarbon dates from bulk sediment or aquatic macrofossils are often imprecise or contaminated by old carbon (SI Appendix, Text), and, as a result, chronologies developed in this way are unreliable.To assess the chronology of shrub expansion and megafauna community turnover in eastern Beringia, we reanalyzed 15 lake sediment records for which AMS ¹⁴C dating of terrestrial macrofossils is available (SI Appendix, Figs. S1 and S6–S9). We developed Bayesian age–depth models for each study site, and compared the results with a new database of published ¹⁴C dates from plant macrofossils, megafauna remains, and arctic ground squirrel middens (Materials and Methods and SI Appendix, Text). In each pollen record, we define the beginning of shrub tundra expansion as the first sustained increase (replicated in three or more consecutive pollen samples) in Betula pollen above pre-15-ka background values, which are typically <5% of the pollen sum (SI Appendix, Fig. S2 and Text). In most cases, this expansion represents an increase to >20% of the pollen sum, and, where they are available, we use pollen influx data to support this timing (SI Appendix, Fig. S3). In some records, Salix pollen increases in abundance before Betula by as much as 1,200 y, and, in these cases, we consider the taxa separately (Fig. 2 and SI Appendix, Figs. S2 and S3). We define the timing of Salix expansion as the first sustained increase in Salix pollen above background values (see above). In most cases, this expansion represents an increase to >15% of the pollen sum. This approach is conservative. It provides minimum ages for the beginning of shrub expansion, as the true increase in shrub pollen above these thresholds is likely to lie between sampling points (i.e., would have an older assigned age). In records with high-resolution sampling [e.g., Birch Lake (19)], this difference is small; however, in most records, sampling resolution is ≤1 pollen spectrum every 10 cm, which may represent >500 y of sediment accumulation (SI Appendix, Table S1). With this approach, we aim to establish whether shrub expansion began prior to turnover in megafauna communities, as predicted by Guthrie (6, 8), or after populations of keystone species collapsed, as suggested by Zimov et al. (2, 29). As these hypotheses predict events in the opposite order, it allows us to assess whether the Late Pleistocene extinction of grazing megafauna species was a response to, or the cause of, steppe-tundra decline.Open in a separate windowFig. 2.(A) North Greenland ice core project (NGRIP) δ¹⁸O record (86). (B) Modeled, calibrated age ranges (shown as probability density functions) for the beginning of Salix (shown in blue when clearly defined) and Betula (shown in gray) expansion from lake sediment records reanalyzed in this study. The medians of calibrated modeled dates are indicated by black crosses. (C) The median of calibrated ¹⁴C age ranges from shrub macrofossils in eastern Beringia. (D) Kernel Density Estimation modeled distributions (mean and 1σ uncertainty) for calibrated ¹⁴C dates from moose in eastern Beringia (sum probability distributions shown in SI Appendix, Fig. S3). (E) Kernel Density Estimation modeled distributions (mean and 1σ uncertainty) for calibrated ¹⁴C dates from horse, bison, and mammoth in eastern Beringia (sum probability distributions shown in SI Appendix, Fig. S3). (F) The median of calibrated ¹⁴C age ranges from arctic ground squirrel middens in eastern Beringia. (G) Periods of human occupation at archaeological sites in the Tanana River Valley, Alaska (70).