Abrupt Holocene climate change as an important factor for human migration in West Greenland

Abstract

West Greenland has had multiple episodes of human colonization and cultural transitions over the past 4,500 y. However, the explanations for these large-scale human migrations are varied, including climatic factors, resistance to adaptation, economic marginalization, mercantile exploration, and hostile neighborhood interactions. Evaluating the potential role of climate change is complicated by the lack of quantitative paleoclimate reconstructions near settlement areas and by the relative stability of Holocene temperature derived from ice cores atop the Greenland ice sheet. Here we present high-resolution records of temperature over the past 5,600 y based on alkenone unsaturation in sediments of two lakes in West Greenland. We find that major temperature changes in the past 4,500 y occurred abruptly (within decades), and were coeval in timing with the archaeological records of settlement and abandonment of the Saqqaq, Dorset, and Norse cultures, which suggests that abrupt temperature changes profoundly impacted human civilization in the region. Temperature variations in West Greenland display an antiphased relationship to temperature changes in Ireland over centennial to millennial timescales, resembling the interannual to multidecadal temperature seesaw associated with the North Atlantic Oscillation.

The Baffin Bay region, comprising West Greenland and the eastern coast of Baffin Island, has a rich and dynamic history, with multiple episodes of human migration and cultural transitions during the mid to late Holocene (1–8) (SI Appendix, Fig. S1). Human occupation in Greenland began with the Saqqaq culture, which arrived ca. 4,500 y B.P. and soon colonized the island (3–5). Beginning ca. 2,800 y B.P., the Saqqaq were replaced by the Dorset culture, which persisted in West Greenland from ca. 2,800–2,200 y B.P., when records of their occupancy end abruptly (3). Archaeological evidence indicates the region remained unpopulated until the arrival of the Norse settlers ca. 1,000 y B.P. (3, 6). The Norse occupied the western settlement (Fig. 1 and SI Appendix, Fig. S1) until the middle of the 14th century (ca. 650 y B.P.) and the eastern settlement until ca. 550 y B.P. (6), whereas the Thule, ancestors of the modern Greenlandic Inuit who arrived two to three centuries after the Norse (7), remain in Greenland to the present day. Human migrations can result from numerous factors, including social change, conflict with neighbors, the search for better hunting grounds, response to shifting resources, economic stress, and climate change (2, 3, 6–8). Although it is difficult to determine how social factors may have influenced migrations, paleoclimate reconstructions offer the opportunity to examine the changing climatic conditions experienced by human populations in West Greenland. The existing lines of evidence for evaluating the impact of climate change on past human populations in West Greenland are generally qualitative and indirect. Greenland ice cores offer quantitative records of temperature from atop the ice sheet (9); however, climate did not change uniformly across Greenland during the late Holocene (10) and quantitative temperature estimates from locations adjacent to and representative of settlement sites are required to assess the influence of climate on observed patterns of human migration.

Fig. 1.

(A) Site map identifying Kangerlussuaq and other locations in West Greenland mentioned in the text. (B) Map of the Søndre Strømfjord region showing locations discussed in the text. (C) Map of the North Atlantic region noting the locations of Kangerlussuaq, Crag Cave (38), and the Bermuda Rise (39). Coloration indicates the modern temperature pattern observed during the positive mode of the North Atlantic Oscillation, where red is warm and blue is cool.

An Alkenone-Based Paleotemperature Record for West Greenland

Here we present a record of temperature variability with decadal-scale resolution for approximately the past 5,600 y from two independently ¹⁴C dated, finely laminated sediment cores from two meromictic lakes, Braya Sø and Lake E (approximately 10 km apart) in Kangerlussuaq, West Greenland (67° 0′ N, 50° 42′ W; Fig. 1 and SI Appendix, Section 1). Our paleoclimate time series are based on measurements of alkenone unsaturation (), a well-established method for reconstructing sea-surface temperatures from marine sediment cores (11) (SI Appendix, Section 2) and whose utility has been demonstrated in lakes from Europe, China, and North America (12–14). The alkenone producers in Braya Sø and Lake E are a newly discovered member of the algal class Haptophyceae (15, 16) and bloom between mid-June and mid-July (Fig. 2A). Air temperature has been shown to exert primary control on lake water temperature in the Søndre Strømfjord region (17). Furthermore, instrumental temperature measurements indicate that Kangerlussuaq air temperature is well correlated with the air temperature of other sites on West Greenland (18) (Pearson correlation coefficients decrease gradually from 0.83 to 0.52 as distance from Kangerlussuaq increases from 130 to 1,200 km for the period 1961–1990 AD). Therefore, Kangerlussuaq lake water temperature primarily reflects air temperature in the Kangerlussuaq area, which is correlated to that of other West Greenland sites as a function of distance. We developed a -temperature calibration for the lakes in Kangerlussuaq (Fig. 2B) by combining a calibration from Braya Sø, based on of filtered alkenones and in situ water temperature, with a previously published lacustrine alkenone calibration from sites containing alkenones of similar molecular composition to those in the Greenland lakes (12, 15) (SI Appendix, Fig. S2). The in situ calibration approach has been successfully employed in North American lakes (14). Our calibration (Fig. 2B; T = 40.8 , r² = 0.96, n = 34), which has a mean standard error of estimation of 1.3 °C (SI Appendix, Section 3), implies that mid-June to mid-July lake water temperature in the Kangerlussuaq region varied by as much as 5.5 °C over the past 5,600 y (Fig. 3A). Alkenone production began approximately 500 y earlier in Braya Sø than in Lake E, and this interval (approximately 6,100–5,600 y B.P.) includes the warmest reconstructed lake water temperatures from either time series (Fig. 3A). Because this time interval is not represented in Lake E, we cannot determine whether this reflects a local period of extreme mid-Holocene warming, if the high values reflect non-temperature-related factors (e.g., nutrients, growth rates) as the haptophyte algae colonized and exploited a new habitat, or if the temperature sensitivity of the alkenones becomes nonlinear at warm temperatures (a possibility that would be undetectable with our calibration approach). We therefore do not interpret this 500 y time period in the Braya Sø record. During the 5,600 y where they overlap, the Braya Sø and Lake E alkenone records show very similar trends on multidecadal to millennial timescales (Fig. 3A; r = 0.53, p < 0.001). After adjusting the two records within 2σ of their calibrated ¹⁴C ages, they were resampled at 20-y intervals and averaged to generate a single record that best estimates past lake water temperature variability in the Kangerlussuaq region of West Greenland (Kanger Stack; Fig. 3B).

Fig. 2.

(A) C₃₇ alkenone flux to Braya Sø lake bottom determined using interval sediment traps. (B) Temperature calibration developed for this study using in situ from Braya Sø water filters collected during summer 2007 (red diamonds) and 2009 (blue squares) and a previously published calibration (12) from Europe (black circles).

Fig. 3.

(A) Alkenone-based lake water temperature reconstruction for Lake E (gray) and Braya Sø (red), Kangerlussuaq, West Greenland. The time series have been visually aligned within the 2σ error of the calibrated ¹⁴C dates. Error bars show standard error of estimation (S.E.) from the calibration and the analytical uncertainty (2σ) (B) The Kanger Stack was developed by resampling and calculating the arithmetic mean of the individual temperature reconstructions from Braya Sø and Lake E at 20-y intervals. Blue shading represents uncertainty from averaging the two records. (C) The loss-on-ignition paleoproductivity record from lake SFL4-1 (21), near Kangerlussuaq. (D) Temperature reconstruction from the Greenland Ice Sheet Project Two ice core, Summit, Greenland (23).

Kangerlussuaq lake water temperatures cooled by approximately 4 °C between 5,600 and 5,000 y B.P., followed by warming of approximately 5.5 °C that culminated between 3,200 and 3,000 y B.P. and then another sharp temperature drop of approximately 5 °C by 2,800 y B.P. (Fig. 3B). The temperature history from alkenone paleothermometry is consistent with previous qualitative temperature inferences for West Greenland from pollen (19, 20) and loss-on-ignition analyses (Fig. 3C) from lake sediments (21) that indicate peak warmth between 4,000 and 3,000 y B.P., followed by Neoglacial cooling ca. 3,000 y B.P. The Greenland Ice Sheet Project Two ice core from Summit (9, 22, 23) (Fig. 1A) also depicts similar trends for millennial-scale temperature variability (Fig. 3D). The large cooling event centered at 2,000 y B.P. in the alkenone record (lake water temperatures dropped by 2–3 °C) corresponds to the greatest Neoglacial advance of inland ice near Kangerlussuaq (24). A cooling trend from 3,000–2,800 y B.P. and the cold interval from 2,200–1,800 y B.P. also coincide with proxy-inferred episodes of increased eolian activity in the region, indicating enhanced windiness and aridity (25). Diatom-inferred lake salinity and lacustrine carbonate δ¹⁸O records (26, 27), plus ¹⁴C-dated paleo-shorelines from Kangerlussuaq (28) document an arid phase from 6,500–4,500 y B.P. followed by an interval of increased moisture from 4,500–3,500 y B.P. These intervals correspond to cooling and warming trends, respectively, in the alkenone temperature data (Fig. 3 A and B). Over the modern instrumental period, correspondence between aridity (humidity) and cold (warm) temperatures in West Greenland is largely due to the effects of the North Atlantic Oscillation (NAO) (18, 29). Although it has been previously suggested that the NAO influenced the climate of West Greenland throughout the mid to late Holocene (30) and there is some evidence (31, 32) for multi-centennial-scale mode dominance of the NAO (periods during which the NAO spends more time in either the negative or positive mode), it remains uncertain whether or not the NAO truly exhibits mode dominance over multicentennial to millennial timescales and we are thus hesitant to ascribe the observed climatic shifts to the NAO.

The History of Human Occupation of West Greenland Related to Climate Change

A comprehensive explanation of the human migration history of West Greenland over the past 4,500 y requires integration of the combined effects of climatic change, environmental degradation, economic stress, social conflict, and a variety of cultural factors. Although such a detailed assessment of West Greenland’s settlement history is beyond the scope of this paper, our study allows evaluation of human migration in West Greenland in the context of Holocene temperature variability. Arriving in Greenland ca. 4,500 y B.P. (3–5), the Saqqaq would have experienced an interval of warmth identified in our temperature reconstruction (Fig. 4A). They survived transient episodes of warming and cooling, especially between 4,100–3,400 y B.P. A cooler interval in the later phase of their occupation at about 3,400 y B.P. coincides with a contraction in the Saqqaq population toward West Greenland (3–5) and a shift from subarctic to arctic conditions in the Disko Bugt region (33) (approximately 250 km north of Kangerlussuaq; Fig. 1A) that likely affected resource availability (34). There is evidence from the Nipisat site near Sisimiut (Fig. 1B) that the Saqqaq developed new adaptive strategies around this time (4, 5). It is possible that climate variability during this period encouraged the diversification of the Saqqaq resource base with, for example, walrus becoming a more important food source. It is uncertain whether local climate conditions, increased contact with other members of the Arctic Small Tool tradition, or other factors are responsible for the adaptations observed in the archaeological record, but with the new strategies, the Saqqaq remained active near Sisimiut (approximately 100 km west of Kangerlussuaq; Fig. 1 A and B) until ca. 2,800 y B.P. (3–5). Perhaps the persistence of warm temperatures in Kangerlussuaq and Sisimiut afforded the Saqqaq adequate access to marine and terrestrial resources (4). Furthermore, large caribou populations in the Sisimiut region would have provided a terrestrial food source as marine resources became more difficult to procure (3). The Saqqaq departure from Sisimiut ca. 2,800 y B.P. is coincident with the culmination of a pronounced cooling trend recognized in our temperature reconstruction (approximately 4 °C in 200 y; Fig. 4A). The cooling that took place during this climate transition was no more abrupt than the transient cooling episodes that took place during the previous approximately 1,500 y of Saqqaq occupation and suggests that the magnitude was more important than the rate of change to the Saqqaq abandonment of the region.

Fig. 4.

(A) Alkenone-based Kangerlussuaq lake water temperature reconstruction (blue curve). Error bars depict the standard error of estimation (S.E.) from the calibration and the analytical uncertainty (2σ) (B) δ¹⁸O record from speleothem CC3, southwestern Ireland (38) (red curve; lower δ¹⁸O values reflect colder temperatures). (C) δ¹⁸O record from Sargasso Sea (39) (black curve, lower δ¹⁸O values reflect warmer temperatures). (D) Difference between normalized time series (Greenland-Ireland) after 21-point smoothing.

The Dorset occupation of West Greenland began ca. 2,800 y B.P. (3) and their tool inventory, which included sledge shoes, soapstone vessels for burning seal fat, and snow knives, suggests they were better adapted to sea-ice hunting than were the Saqqaq (3, 35). The Dorset lifestyle appears therefore well adapted to the colder conditions from 2,800 and 2,000 y B.P. observed in the Kangerlussuaq temperature reconstruction (Fig. 4A). Like the brief warming and cooling episodes experienced by the Saqqaq, the period of Dorset occupation was also characterized by high-amplitude centennial-scale temperature variability, but their continuous occupation in the region indicates that their livelihood strategies allowed them to cope with these temperature fluctuations. These observations prompt the question whether abrupt climate change played a role in the Dorset abandonment of West Greenland. Given the complexity of human behavior, archaeological approaches alone cannot answer this question, but paleoclimatological investigations can help address it. Moros et al. (33) inferred warm sea-surface temperatures and limited sea ice in the Disko Bugt region (Fig. 1A) from 2,000–1,800 y B.P. and suggested that these conditions would have been unfavorable to the Dorset, given that they were predominantly sea-ice hunters. The progressive warming of Kangerlussuaq lake waters to nearly pre-Neoglacial temperatures at this time (2,000–1,800 y B.P.; Fig. 4A) represents a strongly amplified feature of the centennial-scale variability. These temporal data reinforce Moros et al.’s (33) evidence for increased temperature and support the idea that dramatic regional climate change could have greatly impacted the Dorset culture at this time. However, it seems unlikely that there was an extant Dorset population in West Greenland to experience it because archaeological evidence places the latest population at ca. 2,200 y B.P. (3), two centuries prior to the observed warming (33) (Fig. 4A). Moreover, faunal remains from the Dorset Malmquist site, which is located between Kangerlussuaq and Sisimuit (Fig. 1B), are dominated by caribou, indicating that the resource base was diverse and not solely tied to sea-ice hunting (3). In addition, there is an absence of archaeological evidence for a northward migration of the Dorset at this time, which would seem likely if warming temperatures forced them from West Greenland. Thus, the Dorset disappearance may not be directly related to the intense warming from ca. 2,000–1,800 y B.P. Perhaps their disappearance is related to the particularly cold interval observed in the Kangerlussuaq temperature reconstruction beginning ca. 2,200 y B.P. and centered at ca. 2,000 y B.P., except that it is difficult to explain why the impact on Dorset livelihood during this particular cool episode was sufficient to cause the abandonment of the region in contrast to two previous comparable cooling episodes (ca. 2,800 and 2,500 y B.P.; Fig. 4A).

The temperature reconstruction also reveals warming from ca. 1,100 to 850 y B.P., coincident with Norse migration to Greenland. The abrupt temperature decline beginning ca. 850 y B.P. (4 °C in approximately 80 y; Fig. 4A) coupled with the persistence of cooler temperatures until approximately 630 y B.P. yielded a progressively unfavorable climate over several decades with a cumulative adverse effect on the sedentary Norse farming population in West Greenland (36). Thus, a shift toward lower temperatures likely contributed to the abandonment of the western Norse settlement near Nuuk (Fig. 1A) at ca. 650 y B.P., which supports arguments that climatic deterioration played a critical role in the demise of Norse settlements in Greenland (6, 36, 37) (Fig. 1A).

Spatial Patterns of Temperature Change Across the North Atlantic

How did temperature variability in Kangerlussuaq compare with other regions around the North Atlantic? Remarkably, the Kangerlussuaq temperature record shows a strong antiphased relationship at centennial to millennial timescales with the δ¹⁸O record from speleothem CC3 at Crag Cave, southwestern Ireland (38) (Figs. 1C and 4 A and B). The Crag Cave record is not simply temperature (moisture source and precipitation amount may also play a role), but provides a qualitative record of climate variability with higher δ¹⁸O corresponding to warmer conditions (38). Together the records indicate multicentennial to millennial length intervals of anticorrelation between temperature in West Greenland and southwestern Ireland over the past 5,600 y (Fig. 4 A and B). When the normalized records are differenced (Fig. 4D), the intervals 5,200–4,600; 3,000–1,900; and 1,200–0 y B.P. are shown to be characterized by cold Kangerlussuaq temperatures and warm temperatures in southwestern Ireland, whereas the intervals 4,600–3,000 and 1,900–1,600 y B.P. depict warm temperatures in Kangerlussuaq and cold temperatures in southwestern Ireland. Furthermore, cross-spectral analysis reveals that the records are highly coherent (above the 95% confidence level) at periods of 252–278, 211–231, and 122–124 y (SI Appendix, Fig. S7). If such coherence in temperature variability across the North Atlantic were observed at known periods of NAO variability (which are too short to resolve from this dataset), we would likely attribute forcing to the NAO (29). Although there is no clear evidence that the NAO operates at the periods identified from our coherency analysis, multicentennial persistence of NAO mode dominance has previously been suggested from paleoclimate records spanning the Holocene (31) and the past millennium (32). It is possible that the intervals identified in Fig. 4D correspond to periods during which the NAO spent more time in one mode, and we note that a δ¹⁸O-based qualitative sea-surface temperature (SST) record from the Bermuda rise (39) (Fig. 4C) supports this interpretation, depicting millennial-scale warming of western Atlantic SSTs corresponding to cold intervals in West Greenland and warm intervals in southwestern Ireland, and fitting the expected NAO spatial pattern (Figs. 1C and 4 A–C). However, although the spatial temperature patterns are suggestive of the NAO, the available data are insufficient to constrain the observed temperature patterns to NAO forcing. Consideration of volcanic and solar forcing is necessary to explain temporal patterns of climate variability over the past 1,000 y (40) and the influence of these climate drivers likely predate the past millennium. Our results indicate that the ice-free region of West Greenland experienced much greater temperature variability during the Holocene than the top of the Greenland ice sheet, probably as a result of interactions among oceanic, atmospheric, solar, and volcanic forcing, and that climatic changes were important in influencing cultural transitions and human settlement patterns in West Greenland.

Materials and Methods

Sediment cores were sampled at 0.5-cm intervals. Samples were freeze-dried and homogenized by mortar and pestle and extracted with dichloromethane using an ASE200 (Dionex). Total lipid extracts were dried under N₂ gas and quantified by gas chromatography-flame ionization detection (HP6890 Series). was measured with a precision of ± 0.01, using an alkenone standard run once every 10 sample injections, and yielding analytical precision of ±0.4 °C. Alkenones from representative samples were identified by comparison of mass-spectral data with previously reported standards and GC retention times (15). To evaluate the timing of alkenone production, sediment trap material in Braya Sø was collected in 10-d intervals at approximately 20 m water depth during the summer of 2006 using a Technicap PPS 3/3 automated cylindroconical sediment trap (collecting area = 0.125 m²). Material collected in sediment trap bottles was filtered through precombusted (550 °C) Whatman© glass microfiber filters, freeze-dried, and processed as sediment samples (described above), without homogenization. To establish an in situ -temperature calibration, lake water samples from Braya Sø (approximately 1 L each) were collected from various water depths in late-June 2007 and 2009 and filtered through precombusted (550 °C) Whatman© glass microfiber filters, which were freeze-dried and processed as sediment samples (described above), without homogenization. Water temperatures at each depth were determined by direct measurement using a YSI, Inc., sonde.

Total organic carbon for core BS01-01 (E-01) was measured at Brown University (University of Nebraska-Lincoln) on a Carlo-Erba elemental analyzer after acidification of samples with HCl to remove carbonates.