Current Siberian heating is unprecedented during the past seven millennia

Global mean temperature is about 1.2°C above pre-industrial (1850–1900) levels, while Arctic near-surface regions have warmed nearly twice as fast, with the 2011–2020 Arctic mean 0.71°C higher than the preceding decade. Consequences already observed include enhanced Greenland ice loss, record-low sea ice extent, permafrost thawing, and unprecedented Siberian wildfires, with broad implications for natural and human systems. Despite temperature’s central role in the Arctic system, instrumental records in polar regions are often short or incomplete and precisely dated, high-resolution proxies are scarce. Siberia is among the regions experiencing the strongest recent warming; heatwaves have intensified, including a 38°C record inside the Arctic Circle in 2020. This warming accelerates sea-ice loss, permafrost thaw, and extreme wildfires, with cascading ecological and societal impacts and potential feedbacks to the global climate via greenhouse gas release from thawing permafrost. High-resolution Holocene proxy records from Siberia are critically lacking. Many Arctic proxy records come from North America, Scandinavia, and Greenland, leaving the Siberian Arctic underrepresented. Existing archives often have low temporal resolution (e.g., pollen, laminated lake sediments), capture little variability below multi-century scales, and frequently end in the mid-20th century, limiting their usefulness for assessing the recent accelerated warming. Combined with Arctic climate heterogeneity and spatial interpolation issues, this has impeded a robust consensus on the rate and magnitude of industrial-era warming in a Holocene context for Siberia and limits the use of paleoclimate analogs to anticipate future Arctic changes. To address this gap, the authors use annually resolved, summer temperature–sensitive tree-ring records from the Yamal Peninsula (NW Siberia) to reconstruct 7638 years of June–July temperature variability. The reconstruction reveals a multi-millennial Holocene summer cooling trend culminating at the end of the Little Ice Age and its subsequent reversal to an unusually strong industrial-era warming, unprecedented in rate and magnitude over the mid-to-late Holocene in Yamal.

The study situates itself within a context where Siberian high-resolution Holocene climate records are sparse compared to North America, Scandinavia, and Greenland. Many Arctic proxies rely on low-resolution archives (pollen, lacustrine and marine sediments) that smooth variability below ~300-year scales and often terminate in the mid-20th century, precluding assessment of the most recent warming. Due to Arctic climate heterogeneity, reliance on spatial interpolations, and missing modern coverage, previous compilations have not reached consensus on the industrial-era warming’s rate and magnitude in a Holocene perspective for Siberia. Earlier tree-ring width (TRW) datasets often did not preserve multi-millennial trends, whereas some maximum latewood density (MXD) or wood anatomical series provide stronger temperature signals but are rarer and shorter. The Yamal7k record addresses these gaps by providing a highly replicated, annually resolved TRW chronology spanning 7638 years for the Siberian Arctic, allowing robust benchmarking of recent warming against Holocene variability.

Study area and material: Increment cores from living trees (n=186) and subfossil stem cross sections (n=1425) of Larix sibirica were collected along four south-flowing rivers in the southern Yamal Peninsula (67.0°–67.8° N; 69°–71° E), at elevations of ~20–50 m a.s.l., within the permafrost zone. Sampling began in 1982 with 20 field expeditions. Subfossil logs, preserved in river meanders and permafrost soils, were unearthed by erosion and sampled responsibly with permissions. From >3500 available subfossil TRW series of larch, spruce, and birch, ~1900 were cross-dated; 1425 larch series (plus 186 living-tree cores) were included. Spruce, birch, and larch north of 68° N were excluded. If visible, sampling height was 0.3–1.3 m above root collar; otherwise at the lowest point of the log. Chronology characteristics: The continuous chronology spans 5618 BCE to 2019 CE with at least 4 samples per year. The dataset contains 1611 TRW series with mean segment length 142 years (range 41–452). Annual replication ranges from 4 to 187, averaging 30 series, and sample depth exceeds 5 trees in 98.6% of years (7528/7638 years). Chronology development: A two-curve signal-free Regional Curve Standardization (2-curve SF-RCS) was applied to remove age-related growth trends while preserving low-frequency variability. Trees were sorted into growth-rate classes relative to a global RCS curve, and class-specific RCS curves were used. The RCS-adjusted Expressed Population Signal (EPS), adjusted to represent long-timescale uncertainty, exceeds 0.85 over 87.1% of the chronology; average Rbar in 50-year windows is 0.53 (range 0.27–0.80). Climate calibration and reconstruction: The chronology was calibrated against daily air temperatures from the Salekhard meteorological station (66°32′N, 66°32′E; 35 m a.s.l.) available since 1883 CE, targeting June–July (JJ) mean temperature. Multiple reconstruction methods were tested (ordinary least squares regression, variance scaling, multi-frequency band approach, K-fold mean). All reconstructions were highly consistent (inter-method correlations r>0.97, p<0.001). The variance scaling (VS) reconstruction, referred to as Yamal7k, is emphasized. Calibration/verification statistics indicate strong skill: correlation with JJ temperature r=0.75; independent verification RE=0.70 and CE=0.42 for 1883–1951, and RE=0.67 and CE=0.36 for 1952–2019. Decadal variability is reproduced (r=0.92 between 30-year spline-smoothed chronology and instrumental JJ temperatures). The record shows no divergence between TRW and instrumental temperatures. A resampled chronology reducing modern replication remains highly correlated with the original VS reconstruction (r=0.986), confirming robustness. Analytical approach: Time-window analyses (e.g., 170-, 100-, and 30-year windows) excluding intervals with EPS<0.85 assessed trends versus mean temperatures. Extremes were quantified using thresholds based on percentiles (>95th for warm extremes), and return periods were estimated via Generalized Pareto distributions fitted above the 50th percentile of the series.

• The Yamal7k record exhibits a long-term JJ summer cooling trend of −0.08 °C per kyr over 5618 BCE–1850 CE (~0.6 °C total), consistent with Holocene orbital forcing signatures found in lower-resolution proxies, and unlike many TRW records that lack multi-millennial trends.
• Industrial-era warming (1850–2019) abruptly reverses the orbital-scale cooling, producing increases in JJ temperatures with slopes and magnitudes unprecedented for comparable time windows over the mid-to-late Holocene. For 170-, 100-, and 30-year windows ending in 2019, the warming rate and mean temperatures are outliers relative to the past 7638 years.
• Thirty-year climate norms ending at or after 2002 exceed the reconstructed Holocene variability range (>99th percentile). For windows ≥30 years, at least one period after 2013 ranks first in mean temperature among all such windows across the record.
• Warm extremes have become markedly more frequent: during 1920–2019, 27 out of 100 years exceed the 95th percentile of reconstructed JJ temperatures (with 19 of these after 1980). Only three years (1947, 1932, 1971) fall below the 10th percentile, and none are within the coldest 5th percentile over the entire 7638-year record.
• The median and 95th percentile of industrial-era (1851–2019) JJ temperatures increased by 0.53°C (12.43→12.96 °C) and 0.63°C (14.80→15.43 °C), respectively, relative to 5618 BCE–1850 CE.
• The mean JJ temperature for 1920–2019 (13.47°C) exceeds the natural variability with an estimated return period >4850 years when fitting a Generalized Pareto distribution above the 50th percentile. Excluding the last 100 years when fitting the distribution suggests such warmth would have been virtually impossible at any time in the last seven millennia absent climate change.
• For 1850–2019, both the mean JJ temperature (12.83°C) and the warming rate (0.0173 °C yr⁻¹) exceed the range of natural variability in the study region.

By providing an exceptionally long, annually resolved, and well-replicated JJ temperature reconstruction for the Yamal Peninsula, the study establishes that recent Siberian Arctic summer warming is unprecedented over the past ~7.6 millennia in both rate and magnitude. This finding places contemporary Arctic amplification in a robust Holocene context, highlighting that recent conditions depart from the envelope of natural variability inferred for this region. The increase in frequency of warm extremes, coupled with the near disappearance of cold extremes in the last century, aligns with observed environmental impacts in Siberia, including permafrost thaw, sea-ice loss, and escalating wildfire activity. These changes pose significant risks to ecosystems, infrastructure, and communities, and may contribute to global climate feedbacks through enhanced greenhouse gas emissions from thawing permafrost. Although focused on JJ temperatures and without future projections, the reconstruction supports concerns that continued rapid warming could push the system toward a new climate state with more frequent heatwaves and compounding hazards, underscoring the need for adaptation strategies.

The study delivers a 7638-year, annually resolved tree-ring width-based reconstruction of June–July temperatures for the Yamal Peninsula, revealing a Holocene-scale summer cooling trend terminated by an industrial-era warming that is unprecedented in both rate and magnitude. Recent decades exhibit 30-year norms beyond the >99th percentile of reconstructed variability, a sharp rise in warm extremes, and century-scale mean temperatures with return periods exceeding several millennia, indicating that current conditions lie outside the natural variability envelope for this region. These results provide a crucial long-term benchmark for assessing ongoing and future climatic changes in the Siberian Arctic and their associated risks to natural and human systems.

• The reconstruction targets June–July (summer) temperatures only; it does not address other seasons.
• The study focuses on the Yamal Peninsula region, which may limit direct generalizability to other Arctic subregions with different climate dynamics.
• The analysis explicitly does not include projections into the future.
• As with all proxy-based reconstructions, inferences rely on calibration to instrumental data (Salekhard station) and assumptions inherent to dendroclimatological methods, though extensive replication and validation statistics mitigate these concerns.