Climate on Greenland is intimately linked with global heat transport, as Greenland lies between the Labrador Sea and the Nordic Seas, two areas of North Atlantic Deep Water formation (Bullister et al., 2013). The North Atlantic Ocean was relatively warm during the Pliocene (Lawrence et al., 2009). Cooling at the Plio-Pleistocene transition, which manifested during both interglacial and glacial periods at high latitudes, caused Northern Hemisphere ice-sheet growth and altered the global carbon cycle (Herbert et al., 2010; Lawrence et al., 2010). During this interval, tropical and high-latitude temperature became phase-coupled, with synchronous glacial–interglacial changes but differing amplitudes between regions (Herbert et al., 2010; Lawrence et al., 2010). Through the Pleistocene, glacial periods became increasingly cold, whereas temperature between interglacial periods were relatively stable within each region (Herbert et al., 2010; Lawrence et al., 2010). In the North Atlantic region, marine sediment records indicate that mid- to late-Pleistocene interglacial periods exhibited some temperature variability: MIS 9 and 5 were up to ∼ 1 to 2 °C warmer than the Holocene, MIS 11 had similar temperature to the Holocene but was much longer, and MIS 7 and 13 were approximately 1 °C cooler than the Holocene (Cluett et al., 2021; de Vernal and Hillaire-Marcel, 2008; Irvalı et al., 2020). This contrasts with the eastern Arctic, where lake sediment records indicate “super interglacial”, i.e., extreme, warmth during MIS 11, with similar timing to warmth in the Southern Ocean and Antarctica (Melles et al., 2012).

The highest-resolution climate time series for Greenland are derived from ice cores, some of which span both MIS 1 (the Holocene) and parts of MIS 5 (the Last Interglacial, LIG). These time series suggest that mean annual temperature was 2 to 3 °C warmer than modern during peak Holocene warmth (Kobashi et al., 2017; Martin et al., 2024) and 8 ± 4 °C warmer during peak LIG warmth (Dahl-Jensen et al., 2013). Yet these ice-core time series are annual averages for high elevation and likely do not closely reflect summer temperature at the ice sheet margins, which governs ablation and in turn impacts ice-sheet mass balance.

A reanalysis product using spatial and seasonal information from a climate model and temporal information from ice-core proxy time series found that middle Holocene Greenland-wide summers were 3.6 ± 0.4 °C higher than present (Buizert et al., 2018). Peak Holocene summer temperature around the ice-sheet margins, inferred from biomarkers and insect fossils in lacustrine sediment archives, were 2 to 9 °C higher than present from 11.7 to 5.0 ka, with some evidence for greater warming in northern Greenland compared to southern Greenland (Acharya et al., 2026; Axford et al., 2021, 2013, 2017, 2019; Briner et al., 2016; Cluett et al., 2023; McFarlin et al., 2018; Schneider et al., 2024; Thomas et al., 2018). Scattered sediment sections and one extant lake contain pre-Holocene interglacial material with fossils that suggest LIG July temperature was 4.5 to 9.5 °C higher than present (Bennike and Böcher, 1992, 1994; Brodersen and Bennike, 2003; Hedenäs, 1994; Hedenäs and Bennike, 2003; McFarlin et al., 2018) and Plio-Pleistocene summer temperatures 6.5 to 8.5 °C higher than present (Atti et al., 2024; Bennike, 1990; Bennike et al., 2010, 2002, 2023).

The GrIS is additionally sensitive to moisture supply. Ice-sheet model sensitivity tests suggest that the middle Holocene likely experienced greater winter snowfall, which aligns with some proxy evidence interpreted to indicate greater snowfall (Downs et al., 2020; Thomas et al., 2016, 2020). During the middle Holocene, Greenland also experienced more locally-sourced summer precipitation (Balascio et al., 2013; Thomas et al., 2020) and may have had drier summers compared to present (Aebly and Fritz, 2009; Anderson and Leng, 2004). Therefore, there is evidence that from 11.7 to 5.0 ka, summers around the terrestrial margins of Greenland were warmer than present, accompanied by changes in precipitation seasonality, including greater winter snowfall and drier summers.

Together, these records demonstrate moderate to pronounced warmth in Greenland during interglacial periods, yet significant uncertainties remain. Proxies differ in their seasonal sensitivity, variously capturing peak summer conditions (i.e., July temperature), temperature of the months above freezing, or annual mean temperature (Bennike, 1990; Francis et al., 2006; Otiniano et al., 2024; Raberg et al., 2021). These differences complicate direct inter-proxy comparisons of inferred temperature magnitudes. In addition, the impact of polar amplification, which is the phenomenon whereby temperature changes more rapidly at high latitudes compared to low latitudes due to positive feedbacks, including the sea ice-albedo feedback (Bekryaev et al., 2010; Miller et al., 2010), has not been systematically assessed across available time series. Finally, climate during pre-Holocene interglacial periods, including the LIG, MIS 11, and older Plio-Pleistocene warm intervals, remains poorly constrained due to sparse records, limiting our ability to evaluate the duration, magnitude, and spatial extent of Greenland climate during periods when the GrIS was smaller than present.

Understanding the implications of past warmth for the GrIS requires knowledge of past GrIS extent. Northern Hemisphere ice sheets grew large enough to deliver significant ice-rafted debris to the surrounding oceans at 2.7 million years (Ma) (Jansen et al., 2000; Larsen et al., 1994; Raymo et al., 1987, 1989; Shackleton et al., 1984; St. John and Krissek, 2002). The most direct evidence that the GrIS was likely absent one or more times during the Quaternary comes from cosmogenic nuclide measurements in bedrock beneath the GISP2 ice core, which suggests this site must have been glaciated for at least 820 kyr during the past 1.1 Ma (Schaefer et al., 2016). Continuous time series from marine sediment archives suggest that the GrIS persisted through mid- to late-Pleistocene interglacial periods (Hatfield et al., 2016; Reyes et al., 2014). The southern dome of the GrIS was likely smaller during the prolonged but mild MIS11, but appears to have persisted through brief, warmer interglacials (MIS9, MIS5), perhaps because their warm intervals were of insufficient duration to cause substantial retreat (Cluett and Thomas, 2021; de Vernal and Hillaire-Marcel, 2008; Irvalı et al., 2020).

The minimum GrIS extent during the Middle Holocene has been constrained using proglacial-threshold lakes and reworked marine shells deposited in moraines (Briner et al., 2010; Weidick and Bennike, 2007). This evidence suggests that the timing and size of the GrIS minimum Holocene extent varied around the ice sheet but was generally only minimally reduced in size relative to present (Briner and Bennike, 2025; Briner et al., 2014; Cronauer et al., 2016; Larsen et al., 2015; Tabone et al., 2024). The southwestern portion of the GrIS was likely the farthest behind its modern extent during the middle Holocene (Lecavalier et al., 2014), but more constraints from other regions are required to verify this assessment.

The most detailed understanding of both climate and ice sheet extent in Greenland for a previous warm period is for the early and middle Holocene, a period with moderate warmth and moderate, albeit incompletely known, ice-sheet retreat. The greatest number of constraints of pre-Holocene ice-sheet size are from the more widely studied southern region of Greenland. Due to spatial heterogeneity, it is critical to quantify the ice-sheet response in northern Greenland to a range of interglacial conditions, both warmer and longer than the Holocene. It is also critical to quantify climate during pre-Holocene interglacials and assess the spatial expression and mechanisms influencing interglacial warmth. Sub-ice sedimentary archives offer a means to directly constrain the magnitude and timing of past ice-free periods and contain proxies to reconstruct terrestrial environmental conditions near the ice-sheet margins (Bierman et al., 2024a; Bierman et al., 2024b; Christ et al., 2021, 2023; Collins et al., 2025)

This study constrains temperature and precipitation isotope values during the months above freezing for two periods when the ice sheet was smaller than present as recorded in basal sediments at Camp Century: MIS 11 and the Pliocene or Early Pleistocene (Bierman et al., 2024b). We use lipid biomarkers extracted from sub-ice sediments collected in 1966 (Sect. S1 in the Supplement) (Bierman et al., 2024b; Christ et al., 2021; Collins et al., 2025). The biomarkers are distinctive of plants and bacterial communities that exist in extant ice-free areas and therefore had to have been produced under ice-free conditions. To place our findings in a broader spatial context, we also conduct a meta-analysis of published Greenland interglacial temperature and precipitation isotope values and assess the spatial patterns of warmth and water cycle change. This allows us to assess climate conditions during multiple interglacial periods when the GrIS was smaller than present.

The Camp Century sediment core comprises 3.44 m of subglacial sediments recovered at the ice-bed interface beneath the Greenland Ice Sheet in northwestern Greenland (Hansen and Langway, 1966) (77.1667° N, -61.1333° E; Fig. 1A). The sediments are divided into five stratigraphic units (Units 1–5) based on sedimentological characteristics (Bierman et al., 2024b; Collins et al., 2025). The upper units (Units 3–5) yield a last exposure age of 416 ± 38 ka, interpreted to correspond to the MIS 11 interglacial, based on luminescence dating (Christ et al., 2023), whereas the lower unit (Unit 1) was buried between the Early Pleistocene (> 1.4 Ma; infrared-stimulated luminescence) and the Late Pliocene (< 3.2 Ma; ²⁶Al / ¹⁰Be) (Christ et al., 2021). Additional details on study setting, core recovery, sample subdivision, and processing are provided in the Supplement (Sects. S1 and S2.1).

We extracted, purified, and analyzed n-alkanoic acids (plant waxes) and glycerol dialkyl glycerol tetraethers (GDGTs) from the < 64 µm grain size sediment fractions following established protocols (Holtzman et al., 2025; Thomas et al., 2023) (details in Sect. S2 and Bierman et al., 2024b). We used the mean of two high-latitude lacustrine calibrations to infer air temperature of the months above freezing (MAF) from branched GDGTs (brGDGTs) (Otiniano et al., 2024; Raberg et al., 2021). The associated calibration errors are 2.14 and 1 °C (RMSE), respectively, and we propagated calibration uncertainty and analytical error into inferred temperature values (Sect. S3.1, Eqs. S1 and S2; Fig. S1). We used these calibrations since they were derived from high-latitude Arctic lakes (Sect. 3.1). We inferred lake water δ2H values using the C₂₂ n-alkanoic acid in Units 1 and Units 3 to 5 and an estimated apparent fractionation value for aquatic plants in high-latitude lakes, based on compilations of published studies, of -124 ± 5 ‰ (Dion-Kirschner et al., 2020; Gorbey et al., 2022; Hollister et al., 2022) (Eq. S5, Sect. S3.3). We infer leaf water δ2H values from the C₂₈ n-alkanoic acid in all five units using estimated biosynthetic fractionation values for high-latitude terrestrial plants in bioclimate zones similar to northwestern Greenland, also based on literature compilations, of -143 ± 5 ‰ (O'Connor et al., 2020) (Eq. S6, Sect. S3.3).

Because we have independent summer temperature and precipitation δ2H reconstructions, we can estimate the effects of local condensation temperature on summer precipitation δ2H values, providing insights into the vapor δ2H values of the airmasses arriving at the Camp Century site (Holtzman et al., 2025; Thomas et al., 2018) (Sect. S3.4). We calculated the temperature-dependent equilibrium fractionation factor (Clark and Fritz, 1997; Majoube, 1971) using brGDGT-inferred MAF temperature from the same samples as the leaf wax δ2H measurements and then used the fractionation factor and inferred summer precipitation δ2H values to estimate summer vapor δ2H values in each sample (Eqs. S8–S9).

To place the Camp Century temperature and precipitation isotope reconstructions in the context of modern climate, we first estimated modern air temperature and precipitation isotope values at the Camp Century site under ice-free conditions. This elevation-corrected, bedrock-equivalent modern baseline allows direct comparison between the reconstructed climate and climate that would be expected at the Camp Century site today in the absence of the GrIS. The lapse-rate correction accounts only for the influence of elevation on temperature and precipitation isotope values and does not attempt to incorporate other processes that would accompany ice-sheet retreat, such as changes in surface albedo, atmospheric circulation, or moisture transport, although by correcting to values at a location near the modern ice-sheet margin, we may inadvertently account for some of these factors. Consequently, the corrected values should be interpreted as a topographically adjusted reference baseline rather than a complete simulation of the local climate under ice-free conditions. The nearest meteorological station with precipitation isotope observations is at Pituffik (Fig. 1A) (Hersbach et al., 2020; IAEA/WMO, 2015). We lapse-rate adjusted the Pituffik temperature and precipitation isotope values (at 77 m above sea level, a.s.l.) to align with the elevation at which the Camp Century sediments were deposited, a value that we estimated as follows. Although the ice-sheet bed at Camp Century is 530 ± 30 m a.s.l. today (Morlighem et al., 2017), the land surface elevation would be 810 ± 60 m a.s.l. (280 ± 50 m higher than today) following complete removal of GrIS and full isostatic adjustment of the crust (Paxman et al., 2022). Another possibility is that the GrIS was smaller than present but not entirely absent, leading to the possibility that the full isostatic adjustment yields an elevation that is too high. Alternatively, the GrIS margin was inland of, but close to, the Camp Century site. In this case, a forebulge could have caused the elevation to be tens of meters higher than the full isostatic adjustment value (Brandes et al., 2025). This region in northwestern Greenland was influenced by the forebulges of the more massive Innuitian and Laurentide ice sheets (Caron et al., 2018). Given that crustal rebound begins as the GrIS thins and that local forebulge development could increase elevation above the fully compensated value, we therefore use an elevation of 810 ± 60 m as our best estimate of the exposed land surface, with the uncertainty reflecting plausible deviations from full isostatic adjustment and evolution of crustal elevation during ice-sheet retreat.

We adjust the temperature and precipitation isotope values for the months above freezing at Pituffik from its elevation of 77 m a.s.l. to the Camp Century adjusted elevation of 810 m a.s.l., incorporating 60 m of uncertainty into all calculations. We apply lapse rates established for low-elevation, high-latitude sites of ∼ -5 ± 1 °C km⁻¹ for summer months (Fausto et al., 2009) and -46 ± 8 ‰ δ2H km⁻¹ (Koerner and Russell, 1979) (Eqs. S10 and S11). The 810 ± 60 m adjustment therefore equates to -4 °C and -37 ‰, with elevation-related uncertainties of 1 °C and 3 ‰. We report a mean-elevation adjusted temperature and precipitation anomaly relative to the late 20th century for each unit, and fully propagate uncertainties, including variance among Camp Century samples in each unit, interannual variance in modern observations, and uncertainties in analyses, lapse-rate, and elevation (Sects. S3.1, S3.3, and S3.6).

We compiled published Greenland paleoclimate time series for comparison with the Camp Century data, including eight brGDGT, seven ice-core, five chironomid, and six macrofossil records for temperature, and six lake, one marine, and seven ice-core records for precipitation isotope reconstructions. For consistency, we determined the value of each time series relative to the same anomaly period, the late 20th century (Tables S1 and S2). Herein, the anomaly period refers to observations from 1950–2000 CE, obtained from ERA5 via Climate Reanalyzer (Hersbach et al., 2020) except for ice-core sites, where modern temperature estimates were taken directly from datasets reported in the original ice-core studies (Buizert et al., 2018; Kobashi et al., 2017). We define 1950–2000 CE as our modern baseline, and refer to this period as “modern” throughout the rest of the text, as it represents a period with robust observational coverage and aligns with the sampling period for precipitation isotope observations (IAEA/WMO, 2015).

These temperature data were used to identify months with mean temperatures above freezing at each study site. Modern precipitation isotope values were extracted from the Online Isotopes in Precipitation Calculator (OIPC), which provides globally calibrated long-term mean isotope fields based on observational data and spatial interpolation (Bowen, 2017; Bowen et al., 2005; IAEA/WMO, 2015). For each site, we calculated the mean modern precipitation isotope value for the months above freezing to allow direct comparison with proxy-derived precipitation isotope values for the months above freezing.

The biomarkers in the Camp Century samples likely integrate environmental signals over timescales of centuries to a few millennia, although the exact temporal resolution is not precisely constrained and may vary between samples. Moreover, the Camp Century stratigraphy likely represents discontinuous preservation of interglacial conditions and may not capture full interglacial means, nor do the samples necessarily capture peak interglacial conditions. To enable comparison with other proxy data, we therefore calculated millennial-scale mean values of published paleoclimate time series that approximate peak warm conditions in each time series. To determine peak Holocene temperature, we calculated the mean of the warmest two millennia between 11.7 and 5.0 ka in each dataset. For brGDGT time series, we used the same proxy-to-temperature calibration approach as for the Camp Century samples (Sects. 2.2 and S3.1). For chironomid time series, we used the chironomid-July air temperature calibration of Francis et al. (2006), as that was consistently reported by all publications. For precipitation isotope time series, since some datasets do not have independent temperature constraints, we calculated the mean values for 8 to 5 ka, an interval that coincides with peak Holocene warmth on Greenland (Axford et al., 2021; Briner et al., 2016). We used the same plant wax fractionation factor approach to calculate lake and leaf water isotope values for all plant wax time series. Finally, we quantified uncertainties by propagating both the variability within each time series and the errors associated with calibrations. This approach isolates conditions during peak warmth in each time series for comparison with the Camp Century results. For some time series, we know that this interval of peak warmth represents the maximum temperature for a given interglacial period, whereas for other time series that do not necessarily span an entire interglacial period, including Camp Century, the samples provide a minimum estimate of peak warmth.

To compare reconstructions of Greenland temperature based on different proxies, we must understand the seasonal bias of each proxy. brGDGTs reflect MAF temperature, whereas chironomids and plant macrofossils primarily reflect peak summer (July) air temperature (Bennike, 1990; Francis et al., 2006; Otiniano et al., 2024; Raberg et al., 2021). Analysis of modern seasonal temperature relationships at Pituffik Station using ERA5 reanalysis data (Hersbach et al., 2020) shows that although July and MAF air temperatures are strongly and linearly correlated, MAF increases more slowly than July temperature (Fig. S2). ERA5 provides a spatially and temporally continuous estimate of modern climatology by assimilating available meteorological observations (Hersbach et al., 2020). While uncertainties remain in data-sparse regions, ERA5 estimates near Pituffik are better constrained by the availability of long-term meteorological observations compared with the GrIS interior (Chen et al., 2026; Zhang et al., 2021). July temperature captures peak summer warmth, while MAF temperature incorporates all months with mean temperature above 0 °C, including cooler shoulder-season months, and thus responds more weakly to the same forcing than July temperature. Therefore, chironomid or macrofossil-inferred July temperature anomalies of 7.0 °C would be recorded by brGDGTs as a 4.5 °C MAF temperature anomaly, assuming the modern relationship is constant through geological time. This is likely a poor assumption, since the length of the MAF season changes with climate, likely causing the MAF-July temperature relationship to change. We therefore do not convert from one seasonal value to another but note that the seasonal biases in proxies cause brGDGT temperature anomalies to be smaller than chironomid or macrofossil anomalies, even under the same climate conditions.

There are a few locations in Greenland with both precipitation isotope and temperature reconstructions (Acharya et al., 2026; Grootes and Stuiver, 1999; Kobashi et al., 2017; Thomas et al., 2020). To compare vapor isotope values at these locations with those reconstructed at Camp Century, we used the same approach applied to the Camp Century data for both the modern and Holocene observations, except for ice core time series, where we applied the fractionation factor for temperatures below freezing (Ellehoj et al., 2013). The isotope time series at Flower Valley and Secret lakes (Balascio et al., 2013; Lasher et al., 2017) do not have independent temperature time series from the same lake, so we used inferred temperature from nearby sites (Sect. S3.4).

Once we have both the value for the late 20th century anomaly period and the mean value for the published paleoclimate time series, we calculate anomaly values as at Camp Century by calculating the difference between the Holocene, LIG, or Plio-Pleistocene mean values and the late 20th century value. We incorporated uncertainties from both the reconstructed time series and the modern reference dataset (Sect. S3.7, Tables S1 and S2). The anomaly values presented here may differ from those reported in the original studies, as we use ERA5 temperature data from 1950 to 2000 CE and OIPC stable isotope values as the reference values for all sites, whereas the original studies may have used different time periods or sources for the modern anomaly value.

Branched GDGTs are membrane lipids primarily produced by bacteria and are commonly found in soils and lake sediments (Sinninghe Damsté et al., 2000), whereas isoGDGTs are produced mainly by archaea, particularly ammonia-oxidizing Thaumarchaeota in aquatic and marine environments (Sinninghe Damsté et al., 2000). We detected brGDGTS and the isoprenoid GDGTs (isoGDGTs) crenarchaeol (cren) and GDGT-0 (caldarchaeol, cald) in all Camp Century samples. Abundances of the other isoGDGTs (GDGT-1, 2, 3, and crenarchaeol regioisomer) were generally below the detection limit, and the Branched to Isoprenoid Tetraether (BIT) index was > 0.97 for all samples. Camp Century brGDGT distributions are most similar to brGDGTs from modern high-latitude lakes, rather than peats or soils (Figs. S3 and S4) (De Jonge et al., 2015, 2021, 2019; Dearing Crampton-Flood et al., 2020; Ding et al., 2018; Dugerdil et al., 2021; Guo et al., 2020; Halffman et al., 2022; Kusch et al., 2019; Martínez-Sosa et al., 2021; Naafs et al., 2017a, b; Raberg et al., 2024, 2021; Wang et al., 2020; Zhao et al., 2021) so we infer that the brGDGTs in all units in Camp Century were produced in aquatic settings. Independent classification using the BIGMaC algorithm (Martínez-Sosa et al., 2023) also suggests the Camp Century samples are from a lacustrine depositional environment (Table S3), providing additional evidence for a lacustrine source of brGDGT production across all units.

Although it may seem counterintuitive that glacial sediments are dominated by lake-derived GDGTs, we explain this finding as follows: the modern ice-free landscape in northwestern Greenland is dotted with shallow lakes, in some areas covering more than 80 % of the landscape (Walcott-George et al., 2025). It is therefore likely that the organic matter preserved in both the subglacial and proglacial units of the Camp Century sub-ice sediments, which integrate material from across the landscape overridden by the ice sheet, includes organic matter from lakes. The concentration of brGDGTs in modern lake sediments is five-fold higher than that in modern soils (Fig. S10) (Guo et al., 2020; Hollister et al., 2022; Martínez-Sosa et al., 2021). Lake-derived brGDGTs would therefore readily dominate the brGDGT distribution in a mixture of soil and lake sediments.

Lacustrine brGDGT production can be influenced by suboxic conditions, which may alter the distribution of high-abundance compounds and bias reconstructed temperatures by several degrees (Harning et al., 2025; Raberg et al., 2025). The cald/cren ratio in all Camp Century samples, however, was < 10, indicating likely absence of suboxic conditions. We therefore applied two high-latitude lake-based calibrations (Otiniano et al., 2024; Raberg et al., 2021) (Eqs. S1 and S2; Fig. S1) to estimate MAF air temperatures from Camp Century brGDGTs. These calibrations were selected because they were developed using Arctic Lake datasets, where brGDGT distributions exhibit the strongest sensitivity to mean MAF temperature (Otiniano et al., 2024; Raberg et al., 2021).

We follow the PAGES Iso2k framework (Konecky et al., 2020) to structure our isotope proxy interpretation. We first define the measured proxy material and its inferred source (plant origin, seasonality, and source water for n-alkanoic acids), then interpret the controls on plant source-water δ2H values and finally link these to the climate mechanisms influencing precipitation δ2H values.

Unit 1 in the Camp Century sub-ice material is characterized by poorly sorted, unstratified sediments with variable grain size and shape, consistent with basal till deposited subglacially as a single massive unit (Bierman et al., 2024b). The biomarkers preserved in this unit likely pre-date deposition, having been incorporated into the till from glacier-entrained soils, lakes, and vegetation present before glaciation. Given that Unit 1 was deposited > 1.4 Ma (Christ et al., 2021), these biomarkers likely record climate conditions during the late Pliocene to early Pleistocene. brGDGT-inferred MAF temperatures ranged between 4.3 and 5.0 °C (Fig. 2F), with mean reconstructed values 4.1 ± 3.4 °C above present-day conditions (Fig. 3). The δ2H value of lake water inferred from the C₂₂ n-alkanoic acid ranged from -158 ‰ to -147 ‰, while in Unit 1 the δ2H value of leaf water inferred from the C₂₈ n-alkanoic acid ranged from -130 ‰ to -103 ‰ in Unit 1 (Fig. 2D). Mean summer precipitation δ2H values were 13 ± 14 ‰ ²H-enriched relative to modern summer precipitation δ2H values (Fig. 4A). After accounting for the effect of local condensation temperature, summer atmospheric vapor was 19 ± 14 ‰ ²H-enriched relative to the modern value (Fig. 4B). Leaf water was 44 ± 9 ‰ ²H-enriched relative to lake water (Eq. S7, Figs. 2E and 5B).

Unit 2 consists predominantly of ice with sparse dispersed sediments. This unit has not been directly dated but is most likely in stratigraphic order relative to Units 1 and 3 and therefore is between early-Pleistocene (Unit 1) and mid-Pleistocene (Unit 3) age. Biomarkers in Unit 2 may have been produced prior to deposition or could be roughly contemporaneous. brGDGT-inferred temperatures ranged from 4.8 to 6.6 °C, with greater inter-sample variability than observed in the other units (Fig. 2F). Mean MAF temperatures were 4.5 ± 3.1 °C above modern conditions (Fig. 3). We do not infer mean summer precipitation δ2H values from Unit 2, as the wax distributions and isotope values indicate only terrestrial plant waxes are present in this Unit. The δ2H value of leaf water inferred from the C₂₈ n-alkanoic acid ranged from -116 ‰ to -103 ‰ in Unit 2 (Fig. 2D), which were slightly ²H-enriched (5 ± 9 ‰) relative to the leaf water δ2H values in the other four units but are within uncertainty (Eq. S7, Fig. 5A).

Units 3 through 5 contain well-sorted, finer-grained sediments with abundant plant macrofossils indicative of an ice-free environment (Bierman et al., 2024b; Collins et al., 2025). Deposited by low- to high-energy flowing water, these units likely represent deposition by summertime streams flowing across a periglacial landscape. The sediments appear to be in stratigraphic order and have been dated to approximately 416 000 ± 38 000 years ago (Christ et al., 2023). Since the depositional environment was likely a subaerial stream surrounded by tundra (Bierman et al., 2024b), the preserved biomarkers were probably produced in nearby tundra and lake landscapes (see Sect. 3.1: Proxy Systematics) and then transported and deposited at the Camp Century location. brGDGT-inferred MAF temperature ranged from 4.4 to 6.7 in Units 3 to 5. Mean brGDGT-inferred MAF temperature anomalies relative to modern were slightly higher in Unit 3 (5.2 ± 2.4 °C) than in Units 4 (4.5 ± 1.8 °C) and 5 (4.3 ± 2.9 °C), although all are within uncertainty (Fig. 3). The δ2H value of lake water inferred from the C₂₂ n-alkanoic acid ranged from -154 ‰ to -147 ‰ in Units 3 to 5, whereas the δ2H value of leaf water inferred from the C₂₈ n-alkanoic acid ranged from -125 ‰ to -109 ‰ in Units 3 to 5 (Fig. 2D). Mean precipitation δ2H values for the months above freezing were slightly ²H-enriched, but within uncertainty of modern precipitation isotope values, with Unit 3 (14 ± 14 ‰) slightly ²H-depleted relative to Units 4 and 5 (17 ± 14 ‰ and 17 ± 14 ‰, respectively) (Fig. 4A). Summer atmospheric vapor δ2H values were less strongly ²H-enriched relative to modern vapor in Unit 3 (21 ± 14 ‰) than in Units 4 and 5 (22 ± 14 ‰ and 2 3 ± 14 ‰, respectively) (Fig. 4B). Leaf water was ²H-enriched relative to lake water, by a similar amount in all three units (∼ 39 ± 9 ‰), similar to those values in unit 1 (Eq. S7, Figs. 2E and 5B).

MIS 11 summer temperature and moisture sources inferred from the Camp Century sediments complement existing terrestrial and marine temperature reconstructions from the North Atlantic Ocean and from terrestrial archives around Greenland. Because the Camp Century samples we analyzed were deposited at a location that is currently covered by 1387 m of ice, but must have been ice-free during their production, these biomarker-inferred temperature and water cycle data provide constraints on the sensitivity of the GrIS to climate forcing.

Based on the presence of, age constraints on, and abundance of biomarkers in the Camp Century sediments, we can infer that the Camp Century site was ice-free at least twice during the Plio/Pleistocene, once prior to the deposition of the till that forms Unit 1, and another time during MIS11 (Christ et al., 2023). The climate inferred from the Camp Century samples may not describe maximum interglacial conditions, as it is possible that portions of MIS11 are not preserved in the Camp Century stratigraphy, due to lack of deposition or subsequent erosion (Christ et al., 2021, 2023; Collins et al., 2025). For similar reasons, other interglacial periods, including MIS5 and early Pleistocene intervals, may have been ice-free at Camp Century but did not leave a sediment record. Nevertheless, the presence of plant fossils (Christ et al., 2021) and vapor isotope evidence for a greater contribution of local moisture to northwestern Greenland corroborates the inference of reduced ice-sheet extent during MIS 11 and an earlier interval.

These results shed new light on MIS 11 climate on Greenland. Similar summer temperature during the two intervals studied suggest interglacial temperature was relatively stable throughout the Plio-Pleistocene, similar to previous findings (Herbert et al., 2010). During MIS 11, southeastern Greenland experienced moderate summer conditions similar to the Holocene, but that lasted on the order of 30 000 years, which in turn caused dramatic retreat of the southern portions of the GrIS (Cluett and Thomas, 2021; de Vernal and Hillaire-Marcel, 2008; Irvalı et al., 2020; Reyes et al., 2014). Available evidence from Camp Century suggests that northwestern Greenland also experienced MIS11 conditions similar to the Holocene. Therefore, like in southern Greenland, ice-sheet recession during MIS11 in northwestern Greenland may also have been caused by prolonged, rather than extreme, warmth. This interpretation is consistent with ice-sheet modeling showing that sustained moderate summer temperature anomalies over ∼ 16 000 years can produce substantial GrIS mass loss (Robinson et al., 2017).

By 2100 CE, the Arctic will experience mean annual surface temperatures 3.0 to 10.6 °C higher than 1950–2000 CE (Lee et al., 2023). The Camp Century brGDGT-inferred MAF temperature is at the low end of this range (Fig. 3). Modern warming at high latitudes, however, is more pronounced in winter, which means that MAF temperatures will likely increase less than the predicted annual mean (Donohoe and Battisti, 2013; Dwyer et al., 2012). Cloud and albedo feedbacks, however, substantially increase GrIS sensitivity to warming under CMIP6 scenarios (Mostue et al., 2024). The Camp Century sub-ice sediments provide evidence that, due to sustained but moderate warmth during MIS11, the GrIS was > 100 km behind its present margin, plants colonizing the ice-free land. The new temperature and water cycle constraints provided by biomarkers produced and deposited during MIS11 in Camp Century sub-ice sediments suggest that efforts to reduce both the magnitude and the duration of anthropogenic warmth will be important to restraining future GrIS retreat.