Prolonged drying trend coincident with the demise of Norse settlement in southern Greenland.

Declining temperature has been thought to explain the abandonment of Norse settlements, southern Greenland, in the early 15th century, although limited paleoclimate evidence is available from the inner settlement region itself. Here, we reconstruct the temperature and hydroclimate history from lake sediments at a site adjacent to a former Norse farm. We find no substantial temperature changes during the settlement period but rather that the region experienced a persistent drying trend, which peaked in the 16th century. Drier climate would have notably reduced grass production, which was essential for livestock overwintering, and this drying trend is concurrent with a Norse diet shift. We conclude that increasingly dry conditions played a more important role in undermining the viability of the Eastern Settlement than minor temperature changes.

INTRODUCTION

Norse settlers developed the Eastern Settlement on southern Greenland in 985 CE, and other settlers subsequently expanded the settlements on the shores of the fjords to the south and southwest of Narsarsuaq (fig. S1) (, ). The inhabitants relied primarily on raising livestock on cleared pastureland for sustenance (), and the estimated population reached about 2000 people (). Despite the success of the settlements, the region was largely abandoned by the early 15th century (, ). As some previous investigations note, this abandonment could have resulted from multiple issues, including climate change, management failure, economic collapse, or social stratification (, , ). Nevertheless, climate change has long been considered an important contributing factor (). Southern Greenland was always near the limit of agriculture for the Norse settlers, although they had experience in stock-raising and farming in similarly challenging environments in Iceland and Norway (, ). For example, in wintertime, cattle and some sheep and goats had to be kept in the warm dark byres, and by spring, many cattle were too weak to move and the Norse farmers had to carry them out to pasture (). Therefore, it is likely that a changing climate would have imposed considerable stress on the entire Eastern Settlement community ().

Temperature change has often been cited as an explanation for the demise of the Eastern Settlement, on the basis of the notion that the Norse arrived during a relatively warm interval and were able to survive until temperatures declined during the Little Ice Age (). However, most paleotemperature reconstructions from southern Greenland do not have sufficient temporal resolution to capture such a transition, are not directly calibrated to temperature, or are located relatively far from the inner settlement region (–). Furthermore, modern Greenlandic agricultural production is sensitive to rainfall variability (, ), but there is currently little information on hydroclimate during the Norse Period. To address this matter, we report records of past temperature and hydroclimate at a multidecadal (~30-year) resolution, using branched glycerol dialkyl glycerol tetraethers (brGDGTs) and leaf wax hydrogen isotopes (δ2H) in lake sediments, from Lake 578, which is adjacent to a former Norse farm and is only ~9 km from Qassiarsuk (Brattahlíð), where the Norse had some of the largest farms in the Eastern Settlement (fig. S1) (). On the basis of our temperature and hydroclimate reconstructions, we show that there was no abrupt temperature decline around the time when the Norse settlements were abandoned but that summers became increasingly dry throughout the entire Norse occupancy period.

RESULTS AND DISCUSSION

Southern Greenland temperature variability during the Late Holocene

brGDGTs, membrane lipids produced by bacteria (, ), are used to reconstruct temperature (Supplementary Materials). With a 3-year sediment trap experiment in Lake 578, situated in the Eastern Settlement with Norse ruins within the catchment (Supplementary Materials), the MBT′5ME index (methylation of branched tetraethers) () is shown to be significantly correlated with summer water temperature (). This allows us to apply MBT′5ME to a well-dated sediment core (fig. S5) to reconstruct summer water temperature extending back to ~375 CE (Supplementary Materials). Our summer temperature reconstruction shows an overall gradual cooling trend (−0.07°C/100 years) over the entire time span, tracking changes in summer insolation (Fig. 1). Superimposed on the overall cooling trend are several short-lived subcentury scale oscillations.

Fig. 1.

Temperature profiles from southern Greenland and the related driving factors.

(A) Summer insolation anomaly at 61° N (). (B) Arctic summer temperature anomaly based on a synthesis of multiple Arctic records (). (C) Our brGDGT-inferred summer water temperature from Lake 578. Gray shading represents the calibration errors of ±0.52°C (). The dashed line represents the linear trend. Note that the temperature reconstruction is truncated at 7 cm because of the unconstrained “core-top cooling” for the sedimentary MBT′5ME index (Supplementary Materials). (D) Southern Greenland region glacier extent. The solid squares are 10Be-dated moraine age near Narsarsuaq (), and the thick line indicates the readvanced period of the south Greenland ice margin (). (E) Principal component analysis (PCA) scores of the Lake Igaliku chironomid samples (). (F) δ18O values of Scoop Lake chironomids (). (G) Biogenic silica from Lake N14 (). (H) Diatom-based August SSTs from core MD99-2322 from the SE Greenland shelf (). (I) Reconstructed August SSTs from core Rapid 21-COM, the Reykjanes Ridge (). Orange bar indicates the time span of Norse settlement. The red lines are 5-point running mean. See fig. S1 for site locations.

The Lake 578 temperature reconstruction shows a general agreement with other local records and with a pan-Arctic summer temperature reconstruction (), indicating a prominent cooling trend and coinciding with decreasing summer insolation at 61°N (Fig. 1, A to C) (). The Lake 578 record also tracks southern Greenland ice sheet fluctuations, which advanced around 1450 to 1750 CE () (Fig. 1D). One outlet glacier, the Kiagtût Sermiat, reached its late Holocene maximum between approximately 440 and 610 CE (). However, this is considered as an exception and is unlike the general pattern of other ice margins in southern Greenland. It is possible that the Kiagtût Sermiat was influenced by local topographic conditions or internal dynamics and is not representative of regional glacier fluctuations in southern Greenland ().

Unlike the long-term cooling trend shown in our study and from Lake Igaliku in the inner fjords (Fig. 1E) (), two records from the outer coastal region show a general warming trend during the Late Holocene (Fig. 1, F and G) (, ). This is likely due to the coastal influence of the Eastern Greenland Current (EGC) and Irminger Current (IC). The EGC is a cold, low-salinity current that exports sea ice from the Arctic along the eastern coast of Greenland. The IC is a branch of the North Atlantic Current that brings relatively warm water southwestward toward southern Greenland. The two currents mix around Cape Farewell and then move northwestward along the coast of western Greenland, where the sea ice quickly melts (). Diatom-based August sea surface temperatures (SSTs) from marine sediment cores (fig. S1) suggest warming of both the EGC and IC in the past ~2000 years (Fig. 1, H and I) (, ), which is thought to have caused the warming along the outer coast of southern Greenland (, ). However, at more inland parts of the Eastern Settlement, the climate is more continental, the influence of the ocean is reduced, and summer insolation played a more important role, as seen in other Holocene records from Greenland (, ).

It is worth noting that August SSTs of the EGC were anomalously warm during the Medieval period (), and similar warmth is also seen in the two outer fjord records (, ). However, Lake 578 in the inner settlement region shows only moderate warmth at this time, so the Norse settlements did not experience a particularly warm Medieval period. Although there is evidence for a long-term cooling trend in our record, temperature did not decline during the period of the Norse settlement, and there is no evidence of unusually low temperatures around the time when the Eastern Settlement was abandoned (fig. S7). The late 14th century was one of the warmest periods of the entire record.

Southern Greenland hydroclimate reconstruction and the linkage to the North Atlantic Oscillation

To estimate changes in hydroclimate, we measured the deuterium isotopic composition (δ2H) of both long- (nC29) and mid-chain (nC23) n-alkanes (Fig. 2A), which are sourced predominantly from terrestrial higher plants and aquatic plants, respectively (, ). Plant wax δ2H reflects the hydrogen isotope composition of the source water the plant uses to make its leaf waxes with an offset due to biosynthetic fractionation (). Here, we specifically target aquatic and terrestrial wax isotopes that are influenced by the same processes, but the evaporative enrichment of terrestrial plant leaf water allows us to investigate summer relative humidity (ΔRH) with a dual-biomarker model (, ). We note that although the sedimentary leaf wax source attributions are ambiguous in some west Greenland lakes (), in Lake 578, multiple lines of evidence suggest that the mid- and long-chain n-alkanes in sediments are mainly sourced from submerged and terrestrial plants, respectively. For example, the dominant submerged plant, Myriophyllum sibiricum, has abundant C23 n-alkanes. Detailed proxy interpretations are provided in the Supplementary Materials. Because they share the same ultimate water source, both the δ2H of C23 and C29 are generally in accordance with the oxygen isotopes of the DYE-3 ice core (Fig. 2B) ().

Fig. 2.

Hydroclimate profile from Lake 578, southern Greenland.

(A) Leaf wax δ2H of C23 (aquatic) and C29 (terrestrial) n-alkanes from Lake 578. The black and red shading represents 1 σ proxy uncertainty. (B) DYE-3 ice core oxygen isotopes (). The thick violet line is the 5-point running mean. (C) ΔRH reconstruction from Lake 578 based on leaf wax δ2H and the dual-biomarker model (). Blue shading represents the 90% confidence interval (CI), and the thick blue line is the 5-point running mean. (D) The optimum Nuussuaq (NU) ice cap accumulation history and the 10% least complex interval denoted with gray shades (). (E) Concentrations of organic carbon from Lake 578 (Supplementary Materials). Higher values indicate a high organic matter (OM) content. (F) RABD660;670 index (relative absorption band depth centered in 660 to 670 nm) from Lake 578, indicates the intensity of lake productivity (Supplementary Materials). The thick brown line is the 10-point running mean. Orange bar indicates the time span of Norse settlement. VSMOW, Vienna Standard Mean Ocean Water.

Our reconstructed quantitative ΔRH from southern Greenland, normalized over 1950 to 2016 CE, indicates a persistent wet interval during 600 to 950 CE (Fig. 2C), before the arrival of the Norse. After ~950 CE, the Lake 578 record shows a long-term drying trend until the 16th century, becoming relatively stable thereafter. The hydroclimate variability is accompanied by changes in lake productivity (Fig. 2, E and F). Wetter periods are associated with higher concentrations of organic carbon and chlorins, and vice versa. On the basis of observations from modern Arctic lakes, we speculate that this relationship arises from variations in nutrient transport from the catchment to the lake as moisture conditions vary. The drying trend we observe at Lake 578, from ~950 CE to the 16th century, is similar to hydroclimate shifts in west Greenland (Fig. 2D) (). The average snow accumulation of the Nuussuaq ice cap decreased up to 20% during the transition from the Medieval period to the Little Ice Age (). However, accumulation in west Greenland greatly increased since the early 18th century (), while such a trend is not identified in the Lake 578 hydroclimate record. This suggests that the hydroclimatic conditions in west Greenland and southern Greenland have responded differently from anthropogenic warming. The North Atlantic Oscillation (NAO) is largely accountable for hydroclimate variability in southern Greenland (Fig. 3), but this influence does not extend to the location of the Nuussuaq ice cap. The NAO is defined in terms of air pressure differences between Iceland and the Azores, with a positive NAO index indicating anomalously high pressure over the Azores and vice versa (). Although the NAO is most pronounced in the winter, it is a characteristic of all months of the year (). In southern Greenland, precipitation amounts in June, July, and August are significantly correlated with the corresponding monthly NAO index (Fig. 3, A and B). The Greenland Blocking Index (GBI) describes the high-pressure blocking over Greenland and is broadly negatively correlated with the NAO (). Under a positive GBI, the polar jet stream migrates northward, resulting in a negative precipitation anomaly in southern Greenland (). In general, a negative summer NAO phase is associated with reduced precipitation in summer, with the opposite conditions in a positive summer NAO phase, contributing to a wet climate (Fig. 3, A and B). If the same teleconnection pattern prevailed in the past, this suggests that the wet interval (600 to 950 CE) was associated with a persistently positive NAO ().

Fig. 3.

NAO in southern Greenland and its connection to summer precipitation [June, July, and August (JJA)].

(A) Spatial correlation between summer NAO and precipitation from 1950 to 2020; dashed lines include areas of significant correlations (P < 0.005). The precipitation data are based on the CRU TS 4.04 precipitation dataset from the Climate Explorer (climexp.knmi.nl). (B) The correlation between summer NAO and measured precipitation from the Narsarsuaq weather station (1961–2020) (fig. S1B). The red line is the linear fit, and the shaded area represents the 95% CI. (C) Time series of the annual mean and JJA mean NAO index plotted from 1950 to 2020. Brown columns [corresponding to brown squares in (B)] indicate recently observed and documented summer droughts that led to a notable decrease in grass, crop, and hay production (). NAO index in all figures is excerpted from the “NAO (rotated EOF of Z500, 1950-now, CPC)” dataset from the Climate Explorer (climexp.knmi.nl).

It is unclear whether a shift in the NAO caused the prolonged drying trend between ~950 CE and the early 16th century, since there is conflicting evidence based on different NAO reconstructions (, ). The center of the NAO in the past may appear in different locations, making it challenging to assess the real NAO pattern with proxy-inferred NAO reconstructions (). For example, some NAO reconstructions suggest persistent positive NAO mode during the Medieval period, while other studies contradict this notion (, ). Moreover, a North Atlantic zonal wind profile reconstruction does not show a distinct trend in terms of the jet stream position and intensity in the past 1250 years (). Nevertheless, our finding of increasingly arid conditions from southern Greenland throughout the Norse settlement period is robust and supported by multiple proxies.

The impact of drought on the Norse farmers

Compared to the widely held theory that low temperatures led to the demise of the Eastern Settlements, our evidence shows a distinct hydroclimate shift during the Norse period, compared to only moderate temperature fluctuations (Fig. 4, A and B). Although the climate in southern Greenland is not conducive to animal husbandry, according to our reconstructions, the Norse farmers experienced climate conditions in the early years of Eastern Settlement that would have provided quite favorable growing conditions for agriculture in the region. Subsequently, increasingly dry conditions, as indicated by our ΔRH record, would have decreased the available pasturage in the growing season and thus limited fodder yields that were essential for the sustenance of animals during the winter (). This challenge of limited water availability is illustrated by archeological evidence of irrigation channels in Igaliku (, ). In addition, there was a gradual change in the diet of the Norse farmers over time, toward reliance on marine food sources (Fig. 4, C and D). The Norse diet relied primarily on terrestrial sources at the beginning of the settlement era and transitioned to marine-based food sources over time (, ). The prolonged drying trend (Fig. 4B), perhaps exacerbated by warmer temperatures toward the end of the settlement period, is a plausible reason for a decline in the availability of meat from animals raised on Norse farms, forcing the farmers to hunt sea mammals, which was a more dangerous and uncertain activity (). Meanwhile, the increased sea ice during the later settlement period may have hindered their marine harvesting and interfered with connections between the Norse community and Europe (, , ). Last, an inability to manage increasingly drier conditions would have hampered the resilience of the community, possibly leading to social instability and eventual abandonment.

Fig. 4.

The reconstructed climate conditions from southern Greenland and the Norse diet transition.

(A) BrGDGT-inferred summer water temperature (SWT) from southern Greenland. An error bar of ±0.52°C from the calibration () is shown with the gray shading. (B) Leaf wax hydrogen isotope-inferred ΔRH from southern Greenland. Blue shading represents the 90% CI. The thick blue line represents 5-point running mean. (C) Percentage of marine sourced food in the Norse diet (). (D) The ratio of marine to terrestrial mammal bones [number of identified specimens (NISP)] from a Norse farm in Qassiarsuk (). Orange bar indicates the time span of Norse settlement.

Notably, the threat of droughts has been underscored for the local farmers from southern Greenland in the recent decades (). In 2008, drought in southern Greenland caused a 50% reduction in yield of hay and silage, which was characterized as a national problem (). Because of a lack of precipitation, a substantial hay yield decline occurred in the following years of 2010, 2011, 2012, and 2015 (). All these severe drought summers (including 2019) had concurrent negative summer NAO, and they were the most anomalous in the past 70 years (Fig. 3C). While today, such conditions can be ameliorated by importing hay, that option was not available to the Norse settlers, who were increasingly vulnerable to the persistently drier conditions. We acknowledge that the causes of Norse settlement abandonment are complex, and it is difficult to simply attribute them exclusively to climate change. Nevertheless, our results highlight that the hydroclimate changes were tightly tied to the destiny of the Eastern Settlement.

MATERIALS AND METHODS

Field sampling

We collected a 70-cm sediment core from Lake 578 (61°50′ N, 45°37′ W) in July 2016, using a UWITEC (Austria) gravity corer with a percussion hammer. Terrestrial and aquatic plant samples were obtained in July 2017. Lake water samples were collected in July 2018. All samples were shipped back to the University of Massachusetts Amherst. The sediment core was stored in a dark cold room (4°C) until analysis. Plant samples were frozen until analysis. Water samples were stored in a refrigerator until analysis. For more details, see the Supplementary Materials.

Age-depth model

The age-depth model of Lake 578 sediment core is developed on the basis of radiometric dating. The upper 15 cm of sediment core was subsampled at 1-cm resolution in the field. The 210Pb, 214Pb, and 137Cs activity of these 15 samples was measured using Canberra GL2020R Low Energy Germanium Detector at the University of Massachusetts Amherst. Seven discrete terrestrial macrofossils were collected from the sediment core, and radiocarbon analysis was conducted in W.M. Keck Carbon Cycle Accelerator Mass Spectrometer at the University of California, Irvine. Radiocarbon age estimates were calibrated using the “IntCal20” calibration in the R program “BChron.” For more details, see the Supplementary Materials.

Lipid biomarker analysis

Sediments at 1-cm resolution were freeze dried and homogenized. The total lipid extract (TLE) was acquired using a Dionex Accelerated Solvent Extractor 200 with a solvent mixture of dichloromethane and methanol (MeOH) (9:1, v/v). The TLE was further separated into apolar, ketone, and polar fractions with alumina oxide column chromatography.

To measure brGDGTs, the polar fractions were dissolved in hexane/isopropanol (99:1, v/v) and filtered through 0.45-μm polytetrafluoroethylene (PTFE) syringe filters. A known amount of C46 GDGT internal standard was added to each sample to quantify the brGDGT concentrations. Subsequently, all samples were analyzed using an Agilent 1260 high-performance liquid chromatography coupled to an Agilent 6120 Quadrupole mass selective detector with a method () that can differentiate the 5- and 6-methyl isomers. Mass scanning was conducted in selected ion monitoring mode for mass/charge ratios of 1302, 1300, 1298, 1296, 1292, 1050, 1048, 1046, 1036, 1034, 1032, 1022, 1020, 1018, and 744.

To measure leaf wax hydrogen isotopes, all apolar fractions were purified with silver nitrate silica gel chromatography, and the n-alkane concentrations were determined using a gas chromatograph (GC) equipped with a flame ionization detector. As the n-alkane concentrations of the 1-cm interval samples were generally low, we combined every two adjacent samples yielding 35 isotopic measurements. An additional 11 samples from ~600, ~1000, and ~1400 CE were processed following an identical protocol to increase the resolution of the reconstruction in key time intervals. The δ2H of C23, C25, C27, and C29 n-alkanes for a total of 46 samples were measured on Thermo Delta V Advantage isotope ratio mass spectrometer coupled to a Thermo Trace GC Ultra through a GCC III. All samples were measured in triplicate, bracketed by two or three injections of H2 reference gas, with laboratory internal standards injections between each sample and three times at the beginning and end of each sequence, to track and correct intersample drift. Sample δ2H ratios are expressed in per mil (‰) relative to Vienna Standard Mean Ocean Water (VSMOW). The complete methods are detailed in the Supplementary Materials.