Sedimentary biomarkers reaffirm human impacts on northern Beringian ecosystems during the Last Glacial period

Our understanding of the timing of human arrival to the Americas remains fragmented, despite decades of active research and debate. Genetic research has recently led to the ‘Beringian standstill hypothesis’ (BSH), which suggests an isolated group of humans lived somewhere in Beringia for millennia during the Last Glacial, before a subgroup migrated southward into theAmerican continents about 14 ka.Recently published organic geochemical data suggest human presence around Lake E5 on the Alaskan North Slope during the Last Glacial; however, these biomarker proxies, namely faecal sterols and polycyclic aromatic hydrocarbons (PAHs), are relatively novel and require replication tobolster their supportof theBSH.Wepresentnewanalysesof thesebiomarkers in the sediment archiveof Burial Lake (latitude 68°260N, longitude 159°100W m a.s.l.) in northwestern Alaska. Our analyses corroborate that humans were present in Beringia during the Last Glacial and that they likely promoted fire activity. Our data also suggest that humanscoexistedwith IceAgemegafauna formillenniaprior to their eventual extinctionat the endof the Last Glacial. Lastly, we identify fire as an overlooked ecological component of the mammoth steppe ecosystem.

Human arrival to the American continents remains one of the most debated topics in our understanding the global dispersal and proliferation of Homo sapiens (Goebel et al. 2008;Waters 2019). Most researchers agree that Beringia, the landmass that spanned northeastern Asia and northwest North America during the Last Glacial period (herein referring to Marine Isotope Stages 2 and 3), was the initial entryway of this migration to the Americas (West 1996;Hoffecker et al. 2014Hoffecker et al. , 2016, but genetic data and archaeological artifacts provide a fragmented history of human populations in the region. Despite assumptions that human populations could not survive the cold climate of the Glacial Arctic, cut-marked bones found at the Bunge Toll site (Pitulko et al. 2016), and stone tools found at the Yana site complex (Pitulko et al. 2017), in Siberia place humans in western Beringia c. 48 thousand calibrated years before present (cal. ka BP) and 32 cal. ka BP (Pitulko et al. 2016), respectively. In contrast, the earliest unequivocal archaeological evidence in eastern Beringia dates to 14 cal. ka BP (Swan Point; Potter et al. 2013), roughly contemporaneous with the earliest finds in the American continents. This coincidence of timing supports the hypothesis of a 'swift peopling' of eastern Beringia and the American continents c. 14 000 years ago (Goebel et al. 2008). However, cut-marks on animal bones found in the Bluefish Caves equivocally support human presence in eastern Beringia during the Last Glacial (Morlan & Cinq-Mars 1982;Cinq-Mars 1990;Cinq-Mars & Morlan 1999;Morlan 2003), but have been controversial for decades. More recently, stone artifacts in the Bluefish Caves that had remained overlooked for decades were reexamined and found to indicate human presence during the Last Glacial (Bourgeon et al. 2017;Waters 2019). These findings, combined with genetic data, stimulated the proposal of the 'Beringian standstill' hypothesis (BSH), which suggests a small group diverged from East Asians c. 36 000 years ago, moving to isolation in Beringia before eventually moving south into the continent (Tamm et al. 2007;Skoglund & Reich 2016;Mulligan & Szathm ary 2017;Moreno-Mayar et al. 2018;Waters 2019). Importantly, the BSH is built upon earlier genetic analyses that detected differences between Native Americans north and south of the Laurentide Ice Sheet (Szathmary 1993;Bonatto & Salzano 1997) and anthropological inferences from linguistic analyses (Nichols 1990(Nichols , 2008. It has also been bolstered by dental analyses (Scott et al. 2018) and further genetic analyses of ancient DNA and whole genomes (Moreno-Mayar et al. 2018;Sikora et al. 2019). Though the inferences drawn from linguistic and dental analyses support early divergence of the Native American founder group from its Asian ancestors, these analyses do not place humans in Beringia during the Last Glacial. At present, the lished forms of archaeological evidence, such as stone tools and fire pits, still cast uncertainties on the attribution of cut-marks on bones to humans.
The BSH also has important implications for our understanding of Beringian ecosystem dynamics in the Last Glacial and the role of humans in the extinction of Pleistocene megafauna. During the Last Glacial, Beringia was home to the mammoth steppe ecosystem, a productive, diverse grassland that supported and was in turn enhanced by megafauna herbivores including mammoths, bison, horses, caribou, woolly rhinoceros, saiga and muskoxen (Guthrie 2001;Mann et al. 2013). The ecosystem dynamics and demise of the Beringian mammoth steppe and its megafauna are debated (Guthrie 2001;Yurtsev 2001;Zimov et al. 2012;Mann et al. 2013Mann et al. , 2015Sandom et al. 2014). To date, little research has characterized the role that a Standstill population of humans might have had in the demise of this ecosystem without modern counterparts.
Recent palaeolimnological analyses from Lake E5, located in northern Alaska, provide additional biogeochemical evidence supporting the presence of humans in Beringia during the Last Glacial (Vachula et al. 2019). Measurements of charcoal and polycyclic aromatic hydrocarbons (PAHs) in the E5 sediment record indicate increased fire activity from 32 to 19 cal. ka BP, despite natural lightning ignitions being suppressed by a cold glacial climate (Vachula et al. 2019). In many ignition-limited regions, evidence of increased burning has been observed to coincide with human arrival (Mann et al. 2008;Pinter et al. 2011;Argiriadis et al. 2018). Similarly, the arrival of humans to Australia was accompanied by increased fire activity (Rule et al. 2019). This circumstantial evidence for human presence is further supported by faecal sterol data suggesting human faecal contamination at E5. Thus, the Lake E5 record provides evidence of a sustained human presence in eastern Beringia during the Last Glacial (Vachula et al. 2019). The Lake E5 analyses support the BSH but require replication to ensure their validity and to reinforce the determination of human presence prior to the preservation of archaeological artifacts. These biomarkers (PAHs, faecal sterols) are relatively novel, so their reliability must be scrutinized before a consensus can be reached regarding the BSH.
Here, we present new analyses of biomarkers preserved in a lake sediment core from the Brooks Range in northwestern Alaska. We seek to replicate the analyses undertaken on the Lake E5 sediments and determine whether palaeolimnological data can reliably identify human presence in the absence of archaeological artifacts. Though fire was a hallmark of human presence, its ecological role in the glacial ecosystems of Beringia has not been explored (Bliss & Richards 1982;Guthrie 1982Guthrie , 2001. We therefore also assess the role of humans and fire in the mammoth steppe biome that spanned Beringia during the Last Glacial.

Study site
Burial Lake (68°26 0 N, 159°10 0 W; 460 m a.s.l.; Fig. 1) is located in the Noatak River Basin in the Brooks Range in northwestern Alaska (Kurek et al. 2009;Abbott et al. 2010;Dorfman et al. 2015;Finkenbinder et al. 2015). The lake is deep (21.5 m) and small (0.8 km 2 ), with a relatively small watershed (3.3 km 2 ), one outlet stream, and several seasonally ephemeral inlets . Like Lake E5 in north-central Alaska (Vachula et al. 2019), Burial Lake is one of a small number of lakes that persisted during the cold and arid Last Glacial period . As such, sediments from E5 and Burial are valuable environmental archives.

Sediment core and chronology
Sediment cores were collected from several water depths in Burial Lake (5.0, 7.9, 13.2 and 21.5 m), but sedimentary hiatuses associated with the drier glacial period interrupt all of these archives with the exception of the core from the deepest water (Abbott et al. 2010;Finkenbinder et al. 2015). The sediments preserved in this core span the last 37 000 years. The age-depth chronology of this core was established by Finkenbinder et al. (2015Finkenbinder et al. ( , 2018 using the CLAM age modelling package (Blaauw 2010). The chronology is informed by 15 radiocarbondated terrestrial macrofossils (Fig. 2). In this paper, we refer to sedimentological ages as thousands of calibrated years before present (1950 CE = 0 cal. ka BP).

Organic geochemical analyses
Following Vachula et al. (2019), we quantified polycyclic aromatic hydrocarbons (PAHs) and faecal sterols in the Burial Lake sediment record. Briefly, an accelerated solvent extraction system (ASE) was used to extract lipid biomarkers from sediments with dichloromethane (DCM):methanol (MeOH) (9:1 by volume). The total lipid extracts were chromatographically separated into acid (4% acetic acid in ether) and neutral (DCM: isopropanol; 2:1 by volume) fractions with aminopropyl gel columns. The neutral fraction was further divided into hexane, DCM (containing PAHs) and MeOH fractions with silica gel columns. The MeOH fractions (containing sterols) were further purified with alumina columns using hexane:dichloromethane (1:1) and dichloromethane: methanol (1:1) as eluents. Sterol fractions were derivatized with N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) for 2 h at 80°C before quantification. Analytes were identified and quantified with an Agilent 6890N gas chromatograph (Agilent VF-200 ms 60 m capillary col-umn (0.19250 lm)) paired with a 5973N mass spectrometer (GC-MS). The oven temperature initialized at 40°C, increasing by 7°C per min to a maximum of 320°C. PAH and faecal sterol fluxes were calculated by normalizing measured sedimentary concentrations (ng per g of dry sediment) of biomarkers by dry bulk density (g cm À3 ) and accumulation rate (cm a À1 ). Samples were taken from 3cm-thick sediment intervals, except in cases where sediments were not available (mean sample thickness = 2.6 cm). The dry masses of these samples ranged from 1.0 to 4.9 g (mean = 2.5 g). Following Vachula et al. (2019), we measured stigmastanol, a herbivore biomarker, and coprostanol, a human faecal biomarker. We used coprostanol:stigmastanol ratios to identify human faecal contamination in the Burial Lake sediments (D'Anjou et al. 2012;Vachula et al. 2019). We interpret coprostanol:stigmastanol values exceeding 0.18 to indicate human faecal contamination in the mammoth steppe, as this value represents the likely maximum background coprostanol:stigmastanol input to the lake, based on estimated megafauna abundances and faecal sterol assemblages (Bull et al. 2002;van Geel et al. 2008;D'Anjou et al. 2012;Prost et al. 2017;Vachula et al. 2019).

Published palaeoecological data
We compare our organic geochemical analyses with previously published palaeoecological and palaeocli-mate data from Burial Lake and the region. Sedimentary pollen data from nearby lakes, including Feniak (Eisner & Colinvaux 1992) and Niliq (Anderson 1988), as well as from Burial itself (Abbott et al. 2010) have been published but have important caveats (discontinuities, uncertainties). Thus, we compare our record with the pollen data from nearby Kaiyak Lake ( Fig. 1; Anderson 1985) as it is the only pollen record that fully chronologically overlaps with our analyses. Informed by the originally published radiocarbon data (six bulk sediment samples) of Anderson (1985), we remodelled the age-depth chronology of the Kaiyak Lake core using the CLAM age modelling package (Blaauw 2010). Although the reservoir effect makes bulk sediment a less ideal material than terrestrial macrofossils for reliable dating of Arctic sedimentary records (Oswald et al. 2005), Kaiyak Lake generally offers an accurate depiction of vegetation during the Burial hiatus. We discuss these data and their relationships with characterizations of mammoth steppe vegetation using ancient DNA and nematode assemblages (Willerslev 2014). We also compare our data with published chironomid-based palaeotemperature records from Burial and Zagoskin Lakes ( Fig. 1; Kurek et al. 2009). The chironomidbased palaeotemperature record from Burial Lake was updated to the age-depth model presented in Finkenbinder et al. (2018), which is used in this study.  Dyke et al. 2003). Red points show our study site, Burial Lake, as well as the locations of archaeological finds and other palaeoecological records referred to in the text.

Results
Fluxes of PAHs and coprostanol in the Burial Lake sediment core document increased burning and more pronounced human presence during the Last Glacial compared to the deglaciation and Holocene (Fig. 3). The fluxes of PAHs were greatest from 34 to 16 cal. ka BP (0.0-2.6, mean = 1.2 ng cm À2 a À1 ) and lowest from 16 to 0 cal. ka BP (0.1-1.4, mean = 0.6 ng cm À2 a À1 ), as were those of coprostanol (34 to 16 cal. ka BP: 0.0-0.5, mean = 0.1 ng cm À2 a À1 ; 16 to 0 cal. ka BP: 0-0.2, mean = 0.1 ng cm À2 a À1 ). In contrast, fluxes of a herbivore biomarker, stigmastanol, were relatively stable throughout the last 34 000 years, although there are notably low values during the Last Glacial Maximum (c. 24 to 20 cal. ka BP; Fig. 3). The ratio of coprostanol:stigmastanol, an indicator of human faecal presence in the Beringian mammoth steppe (Vachula et al. 2019), was consistently greater than the background input ratio (0.18) from 34 to 12 cal. ka BP (Fig. 4).

Discussion
Evidence supporting human presence in eastern Beringia during the Last Glacial Organic geochemical analyses of Burial Lake sediments demonstrate the presence of humans in eastern Beringia during the Last Glacial, supporting the Beringian standstill hypothesis (BSH) and previous palaeoecological records from Lake E5. Increased fluxes of PAHs from 34 to 16 cal. ka BP indicate increased fire activity in eastern Beringian landscapes during the Last Glacial (Fig. 3). Although climate was more arid during the Last Glacial, as evidenced by Burial Lake PAH levels ( Fig. 5; Abbott et al. 2010), natural lightning ignitions were likely less frequent in eastern Beringia during the Last Glacial due to decreased convective energy available in the atmosphere (Vachula et al. 2019). Decreased natural ignitions during the Last Glacial are difficult to reconcile with evidence of increased burning. Though increased aridity and more readily flammable vegetation assemblages (i.e. mammoth steppe) might have promoted burning, human ignitions offer an alternative explanation for the magnitude of elevated fire activity. Indeed, during the same period of elevated fire activity (34 to 16 cal. ka BP), the fluxes of coprostanol, a human faecal biomarker, are greatest, and coprostanol:stigmastanol values consistently exceed the 0.18 threshold value indicative of human faecal contamination (Vachula et al. 2019). Taken together, these data suggest that humans played an integral role in fire regimes near Burial Lake. The timing of this inferred human presence agrees with genetically-inferred timing of the initial divergence of Ancient Beringians from East Asians (Tamm et al. To conclusively support the BSH, further research is needed to confirm that the humans evident in the Burial Lake sediment record are indeed the ancestors of Native Americans. Our interpretation of faecal sterol ratios as indicative of human presence in eastern Beringia is grounded by current understanding of the typical sterol assemblages in the faeces of different animals and the abundance of These proxies indicate there was elevated biomass burning and more pronounced human presence during the Last Glacial. Fluxes of herbivore fecal sterols, stigmastanol (C), suggest herbivore populations were relatively stable during the last 34 000 years. The age-depth model of the Burial Lake sediment core, with white circles showing accepted radiocarbon dates and crosses depicting rejected dates (Finkenbinder et al. , 2018. different animals in the mammoth steppe ecosystem of eastern Beringia. Following Vachula et al. (2019), we assume that coprostanol:stigmastanol values exceeding a background input ratio of 0.18 indicate human faecal contamination of the sediment. This threshold value was determined by estimating the likely 'natural' faecal sterol assemblage of sediments that would result from faecal inputs of fauna present in eastern Beringia during the Last Glacial. Namely, 94% of faunal biomass in this region consisted of large herbivores including mammoths (49%), horses (23%) and bison (22%; Mann et al. 2013). Whereas human faeces has coprostanol:stigmastanol values of 2.22 to 5.5 (Bull et al. 2002;Prost et al. 2017), the values expected in the faeces of mammoths, horses and bison are 0.18, 0.11 and 0.11, respectively (van Geel et al. 2008;Prost et al. 2017;Vachula et al. 2019). These values are derived from measurements of coprolites and modern faecal samples and could have varied in the past and with changes in diet or gut physiology. Nonetheless, human faeces have coprostanol:stigmastanol values that are at least an order of magnitude greater than those of the herbivores dominating faunal assemblages of the mammoth steppe, justifying our interpretation of coprostanol:stigmastanol values exceeding 0.18 to indicate human faecal contamination. Other omnivores (e.g. pigs, opossums) could also contribute coprostanol to the sedimentary coprostanol: stigmastanol values (Leeming et al. 1996), which would confound the identification of human faecal contamination. However, palaeontological data indicate that omnivore biomass was quite limited in the mammoth steppe (Mann et al. 2013). Overall, it is unlikely that sedimentary coprostanol:stigmastanol ratios could have exceeded 0.18 during the Last Glacial without human faecal input.
Although we interpret our faecal sterol data to indicate human presence in eastern Beringia during the Last Glacial, we concede that there are alternative explanations for our interpretation of human impacts on fire. Namely, the elevated fire activity from 34 to 16 cal. ka BP, as inferred from increased fluxes of PAHs (Fig. 3), could result from natural climatic forcing. Climate was significantly more arid during the Last Glacial in eastern Beringia , suggesting that elevated fire activity could have resulted from climate rather than human activity. Colder glacial temperatures, however, ought to decrease biomass burning (Hu et al. 2010;Young et al. 2017), thereby bolstering the case for anthropogenic fire activity. Biomass burning proxies cannot distinguish anthropogenic and natural fire activity, complicating the conclusive interpretation of our data. However, our faecal sterol data, in conjunction with the growing body of interdisciplinary research supporting the BSH (Nichols 1990(Nichols , 2008Bonatto & Salzano 1997;Tamm et al. 2007 , provide compelling support for human presence in eastern Beringia during the Last Glacial, and it is certainly possible that their activity promoted more widespread fire activity. However, this anthropogenic burning may be superimposed upon naturally elevated fire activity. In this way, further research is needed to more thoroughly dissect the fire dynamics of the mammoth steppe in eastern Beringia. Though key differences exist between Burial Lake and Lake E5, our proxy results from the Burial Lake sediment core generally agree with those from Lake E5 and support human presence in eastern Beringia during the Last Glacial. Vachula et al. (2019) found elevated charcoal accumulation rates and PAH fluxes from 32 to Fig. 4. Comparison of palaeolimnological data from Burial Lake (top) and Lake E5 (bottom). PAH fluxes in both lakes (A and E) indicate elevated biomass burning during the Last Glacial period, coincident with increased fluxes of the human fecal biomarker, coprostanol (B and F). Stigmastanol fluxes in the Burial Lake record (C) are consistent throughout the last 34 000 years, relative to those in Lake E5 (G). Coprostanol:stigmastanol ratios (D and H) consistently exceed the 0.18 threshold value (orange dashed line) in both lakes during the Glacial period, indicating prolonged human fecal contamination. Overall, the consistency of these results from two disparate lakes in northern Beringia suggests humans were present in Beringia during the Last Glacial and promoted biomass burning.
19 cal. ka BP at Lake E5 (Fig. 4). Our PAH results agree and indicate there was increased burning from 34 to 16 cal. ka BP (Fig. 4), though there is an absence of charcoal in Burial Lake sediments. Although the absence of charcoal might be due to the absence of wood fuels, experimental analyses indicate that non-wood tundra  (Clark et al. 2009). Burial Lake levels (B) fluctuated in response to climate and basin fill (Finkenbinder et al. , 2018. Chironomid-based palaeotemperature records from Zagoskin (C) and Burial Lakes (D) indicate lower July temperatures during the Last Glacial. Sample-specific prediction errors are marked as black lines. Elevated fluxes of PAHs in Burial Lake (E) coincide with palynological evidence for the mammoth steppe from Kaiyak Lake (F; green = trees/shrubs; yellow = upland herbs; cyan = aquatic vascular plants; blue = algal species; brown = terrestrial vascular cryptogams; gray = unidentified palynmorphs) and Burial Lake (G). Elevated fluxes of coprostanol (H) indicate human presence near Burial Lake during the Last Glacial period. Fluxes of stigmastanol (I) indicate consistent presence of herbivores near Burial Lake. Coprostanol:stigmastanol ratios (J) consistently exceed the 0.18 threshold value in both lakes during the Last Glacial period, indicating prolonged human fecal contamination. Red vertical bar depicts the timing of Ice Age megafauna in northeastern Beringia.
plants (graminoids and forbs) produce nearly as much charcoal per fuel mass as do shrubs (Pereboom et al. 2020). Although the abundance of wood fuel in the mammoth steppe might be disputed (Willerslev 2014), it is unlikely to diminish the reliability of sedimentary charcoal as a palaeofire proxy. Like E5, Burial Lake would have experienced suppressed lightning ignition frequency during the Last Glacial (Vachula et al. 2019). The absence of charcoal in the Last Glacial Burial Lake sediments  coincident with increased fluxes of PAHs could indicate more regional vs. local (i.e. catchment) fire dynamics than in the case of Lake E5. Indeed, PAHs can be transported long distances in the atmosphere (Halsall et al. 2001). Moreover, recent work characterizing the source areas of sedimentary charcoal in the Alaskan tundra suggest that charcoal records in this region have a significant extralocal and regional fire history signal (Vachula et al. 2020). Previous work indicates that during the Last Glacial, significant inputs of dust to Burial Lake were likely sourced from exposed continental shelves composing the Bering Land Bridge , suggesting PAHs could have been sourced from this region as well. Notably, the Bering Land Bridge itself has been proposed as a mesic refugium for early humans in Beringia (Hoffecker et al. 2014), so it is possible the PAH signal reflects a population living on the Bering Land Bridge but visiting Burial Lake and its surroundings regularly. Dorfman et al. (2015) also identified a dust source change after 15 cal. ka BP indicative of a more local source of sediments, which could be true for PAHs as well. Notably, our PAH and faecal sterol data from Burial Lake suggest greater fire activity and influx of faecal matter (human and herbivore) during the deglaciation and Holocene than do those from Lake E5 (Fig. 4). However, the Noatak Basin may simply have experienced more frequent burning throughout the last 34 000 years than did E5. Historical records indicate the Noatak Basin experiences more fire activity than the central Brooks Range foothills (E5) due to more extreme fire weather conditions (humidity, temperature, precipitation, andwind speeds;French et al. 2015). Though it is difficult to determine how these variables might have varied between these two sites over the last 34 000 years, this agreement between modern observations and the palaeofire records suggests the vegetation around Burial may have been more susceptible to burning than that around E5. Similarly, herbivores and humans may have had a more pronounced presence near Burial Lake relative to Lake E5 during the deglaciation and Holocene. Elevated fluxes of stigmastanol and sitosterol in Burial Lake from 15 cal. ka BP to present suggest dense herbivore populations persisted during the Holocene (e.g. caribou) whereas they decreased near E5. Notably, fluxes of stigmastanol were low during the Last Glacial Maximum (Fig. 3) at Burial Lake, coincident with palaeozoological evidence for declines in megaher-bivore populations during this interval (Matthews 1982;Sher et al. 2005;Mann et al. 2015). Thus, despite differences between Burial Lake and Lake E5 data, proxy results from both lakes support human presence in eastern Beringia during the Last Glacial.
Additionally, although the general trends of these faecal biomarkers and fire indicators in the two lakes agree, the total fluxes of PAHs and faecal sterols differ between the two lakes, with Burial Lake having lower fluxes by an order of magnitude (Fig. 4). These discrepancies are difficult to resolve. Burial Lake is approximately eight times larger in area than Lake E5, so differences in the efficiency of sediment transport and focusing could explain the lower fluxes. Similarly, preservation and diagenesis differences between the two lakes may add to the discrepancy (Biache & Philp 2013). Sedimentary organic matter content, and its effects on biomarker preservation, could explain some of this discrepancy. Indeed, percent weight of organic carbon in sediments from Lake E5 (6.2-12.7%; mean = 6.2%) were generally greater than in those from Burial Lake (1.2-7.8%; mean = 3.6%), though not sufficiently so to completely resolve this discrepancy. Though it is difficult to determine the cause of discrepancies of PAH and faecal sterols fluxes between Lake E5 and Burial Lake, it is promising that both records support human faecal input and increased burning during the Last Glacial. Vachula et al. (2019) posited that Lake E5 might have been one of the few sources of fresh water for megafauna in arid, glacial Beringia, and so might have attracted megafauna and humans alike. Indeed, such an Arctic glacial oasis might have similarly attracted and concentrated megafauna within the watershed of Burial Lake.
Our approach highlights the utility of palaeolimnological approaches as a means of inferring human presence in the absence of archaeological artifacts. The juxtaposition of the swift peopling and Beringian standstill hypotheses exemplifies disagreement between the fields of genetic anthropology and archaeology. At the core of this disagreement is the paucity of preserved material culture (e.g. stone or bone tools, fire pits) in Beringia during the Last Glacial, despite the finds at the Bluefish Caves (Morlan & Cinq-Mars 1982;Cinq-Mars & Morlan 1999;Morlan 2003;Bourgeon et al. 2017). However, the rarityof these artifacts ought to be expected as the deglaciation in Beringia was accompanied by landscape burial by melting (and occasionally temporary expansion) of alpine glaciers and the Laurentide Ice Sheet, as well as subsequent sea level inundation, which only compounds the difficulty of archaeological artifact recovery in remote permafrost regions (Hoffecker & Frederick 1996;Hoffecker et al. 2014). In light of these practical realities, we suggest that the geochemicalarchaeological approaches undertaken with the E5 and Burial Lake sediments may represent a pragmatic means of reconciling these two hypotheses. Further, these approaches highlight the utility of palaeolimnological records as complementary perspectives to traditional archaeological and genetic research seeking to understand human migrations of the past. However, more work is needed to properly benchmark and test the accuracy of the geochemical proxies presented in this study, preferably at sites where the archaeological record is better constrained.
Fire's role in the maintenance and productivity paradox of the mammoth steppe The ecology of the mammoth steppe has puzzled palaeoecologists for decades, in particular the great abundance and diversity of megafauna and vegetation of the mammoth steppe relative to the low productivity expected of high latitude ecosystems in the Last Glacial, a problem termed the 'productivity paradox' (Vereshchagin & Baryshnikov 1982;Zimov et al. 1995Zimov et al. , 2012Yurtsev 2001;Bradshaw et al. 2003;Willerslev 2014). In the same areas today, tundra ecosystems support less diverse fauna and are less productive despite experiencing warmer and wetter interglacial climatic conditions. Palaeoecologists have offered several explanations for this paradox and the maintenance of the mammoth steppe. Aridity and decreased cloud cover associated with a more continental and cooler climate are thought to have promoted dry, well-drained and productive soils (Vereshchagin & Baryshnikov 1982;Guthrie 2001;Yurtsev 2001). Megafauna are also thought to have helped maintain the mammoth steppe by grazing and trampling shrubs, and to have promoted productivity by increasing nutrient cycling efficiency (Zhu et al. 2018). However, analyses of ancient DNA and nematode assemblages suggest that the woody component of mammoth steppe vegetation may be more significant than inferred from pollen records (Willerslev 2014). Clearly, further research is needed to better understand the ecology of the mammoth steppe.
Fire may also have promoted the productivity of the mammoth steppe by enhancing rates of nutrient cycling (most likely nitrogen) and helped to maintain the steppe by reducing plant litter accumulation and the growth of woody plants (Guthrie 2001), but little research has explored it as a possibility. The development of quantitative palaeofire reconstructions using sedimentary charcoal (e.g. Patterson et al. 1987;Clark 1988;Clark & Royall 1995, 1996 occurred at the same time as much of the work initially characterizing Beringian vegetation during the Last Glacial (e.g. Anderson 1985Anderson , 1988Eisner & Colinvaux 1990, 1992Anderson et al. 1994). Consequently, modern fire reconstruction methods and theory likely had not yet become part of the standard palaeoecological toolbox. For example, Eisner & Colinvaux (1990) noted elevated charcoal in the Last Glacial sediments of Ahaliorak Lake in Arctic Alaska but interpreted it as an indicator of aridity.
Our data suggest that fire and herbivory could have interacted to maintain the mammoth steppe and promote its productivity. Though it is difficult to decipher the dynamics of this no-analogue ecosystem, fire and herbivory have similar impacts on many modern ecosystems (Bond & Keeley 2005). For example, fire and herbivory partly determine the relative amounts of grassy and woody vegetation in savannah ecosystems (Van Langevelde et al. 2003;Smit et al. 2010;Holdo et al. 2012), and they may have done the same to limit shrub expansion in the herbaceous mammoth steppe (Zimov et al. 1995;Guthrie 2001;Blinnikov et al. 2011). However, ancient DNA analyses suggest that the mammoth steppe did have a significant shrub component (Willerslev 2014). Herbivore grazing promotes N enrichment in savannahs (Augustine et al. 2003) and reduces the losses of N to combustion in tallgrass prairie (Hobbs et al. 1991). Similarly, megafauna grazing is thought to have promoted the productivity of the mammoth steppe (Guthrie 1982;Schwartz-Narbonne et al. 2019). The role of fire in promoting ecosystem productivity is less clear (Pausas & Ribeiro 2013). Though fire volatilizes N in prairies, grasslands, rangelands and heathlands (Hobbs et al. 1991;Adams et al. 1994;Blair 1997;Snyman 2003;Henry et al. 2006), it promotes P availability and productivity in the same ecosystems (Butler et al. 2018). In shrub-grasslands and shrub steppes, grazing moderates fire activity by increasing fuel moisture, decreasing fuel abundance and decreasing fuel continuity (Davies et al. 2015(Davies et al. , 2016(Davies et al. , 2017. Though our data do not facilitate a complete analysis of the role of fire in the mammoth steppe, they demonstrate that fire did affect the ecosystem for millennia and highlight a topic needing further research.

The demise of the mammoth steppe and Ice Age megafauna in Beringia
The relative roles of humans and climate in the demise of the mammoth steppe and its megafauna are debated in the literature and are thought to have varied by region (Brook & Bowman 2002;Johnson 2002;Miller et al. 2005;Gill et al. 2009;Lorenzen et al. 2011;Sandom et al. 2014). In northeastern Beringia, climate change and subsequent landscape changes and habitat fragmentation are thought to have caused the extinction of megafauna (Mann et al. 2013(Mann et al. , 2015, though this conclusion rests in part upon the assumption of swift peopling of the region at c. 13.5 cal. ka BP (though other sites suggest an earlier arrival to northeastern Beringia). This stands in contrast to global analyses indicating humans dominated climatic drivers of extinction (Sandom et al. 2014). To date, the timing of human arrival has not been reconciled between the date proposed by the BSH and the date of the megafauna extinctions in eastern Beringia.
Our data, in conjunction with published palaeontological and genetic data, suggest that humans could have influenced faunal population dynamics during the Last Glacial, but their impacts on the ultimate extinction agent of the mammoth steppe and its megafauna during the deglaciation were secondary to climate-driven vegetation changes. Our data place humans in eastern Beringia as early as 34 cal. ka BP, in agreement with the genetically inferred separation of the Ancient Beringian population c. 36 000 years ago (Raghavan et al. 2015;Moreno-Mayar et al. 2018). This suggests a long period of interaction between humans and megafauna, which would be expected to allow for coevolution and less severe extinctions (Sandom et al. 2014). Genetic evidence suggests that Beringian bison experienced precipitous population declines beginning 37 cal. ka BP (Shapiro et al. 2004) and that there was a local extinction of brown bears in Beringia from 35 to 21 ka (Barnes et al. 2002). The timings of these faunal population changes are conspicuous in light of our data and the BSH. Indeed, humans may have influenced bison populations via hunting and may have driven out omnivorous brown bear populations (Matheus 1995). However, herbivore megafauna extinctions in northeastern Beringia did not occur until the deglaciation, with mammoths disappearing 13.8 cal. ka BP and horses and bison remaining until 12.5 cal. ka BP (Mann et al. 2013). These extinctions are coincident with distinct increases in summer temperatures in eastern Beringia (Fig. 5), as inferred from chironomid assemblages (Kurek et al. 2009), as well as decreased aridity, as inferred from increased proportions of algal and aquatic vascular plant pollen and Burial Lake levels (Fig. 5). Geomorphological evidence for climate-driven peatland establishment and alluviation in flood-plains from 14 to 12.8 ka are thought to have fragmented habitats and caused the ultimate demise of these megafauna (Mann et al. 2010(Mann et al. , 2015. Importantly, the timing of these climatic changes coincides with the ages of archaeological finds supporting the swift peopling hypothesis, which likely relate to human population expansion. It is therefore difficult to determine if humans or climate caused the ultimate demise of megafauna in Beringia; however, the extended period of cohabitation during the Last Glacial suggests humans, despite having potential impacts on megafauna populations during the Last Glacial, were secondary to climate.

Conclusions
Our analyses of biomarkers (faecal sterols and polycyclic aromatic hydrocarbons (PAHs)) preserved in the Burial Lake sediment core corroborate previous analyses of Lake E5 sediments. These analyses support the presence of humans in Beringia during the Last Glacial and suggest that they promoted fire activity. The timing of human presence evident in the Burial Lake and Lake E5 sediment records agrees with the Beringian Standstill Hypothesis. However, further research is required to determine if the human population identified by our analyses was indeed the Standstill population. Our data show that Ice Age megafauna and humans coexisted for millennia, suggesting that humans were not the primary cause of the extinctions that occurred at the end of the Last Glacial. Lastly, we show that the role of fire in the mammoth steppe has been overlooked in the palaeoecological literature, but that it may have been an important agent in the maintenance of this puzzling ecosystem's structure and productivity.
Acknowledgements. -We thank F. Husain and N. O'Mara for sampling assistance. The research was supported by NSF grant PLR-1503846 to Y. Huang and J. M .Russell, NSF grant ARC-0909545 to M. B. Abbott, a grant to Y. Huang by the National Park Service Shared Beringia Heritage program, Department of the Interior P18AC00556, and a graduate student fellowship granted to R. S. Vachula by the Institute at Brown for Environment and Society, Brown University. We thank J. Hoffecker for his helpful comments and improvements to the manuscript.
Author contributions. -MBA and MSF collected samples. RSV analysed samples. RSV, MBA and YH designed the project. All authors contributed to the interpretation of data. RSV and JMR drafted the manuscript, which was reviewed and edited by all authors.
Data availability statement. -The data that support the findings of this study are available from the corresponding author upon reasonable request. They will also be made available for download from the NOAA's World Data Service for Paleoclimatology.