Alpine permafrost could account for a quarter of thawed carbon based on Plio-Pleistocene paleoclimate analogue

The study addresses large uncertainties in the magnitude and even sign of the permafrost–climate feedback by determining how permafrost carbon responds under warmer-than-present conditions. It focuses on whether alpine or circumarctic permafrost is more vulnerable to warming and how much permafrost carbon could thaw in a near-future climate analogous to the mid-Pliocene Warm Period (mPWP, 3.3–3.0 Ma) when atmospheric CO2 was ~400 ppm. The authors reconstruct Plio–Pleistocene temperatures from lacustrine sediments on the Tibetan Plateau (Earth’s largest alpine permafrost region) using multiple proxies including carbonate clumped isotopes to infer MAAT and evaluate permafrost presence. They integrate these reconstructions with PlioMIP2 climate model outputs to estimate global permafrost stability and potentially affected soil organic carbon (SOC) in a mPWP-like climate, thereby assessing the relative future vulnerability of alpine versus circumarctic permafrost carbon stocks.

Prior work shows permafrost underlies <20% of land but stores ~1500 Pg organic carbon (~50% of global SOC), with most modern permafrost in circumarctic regions and substantial alpine permafrost on the Tibetan Plateau (~160 Pg SOC). Observations indicate rapid permafrost warming and thaw, potentially releasing CO2, CH4, and N2O, yet models disagree on the permafrost–climate feedback magnitude and even sign. The mPWP is a widely used analog for near-future climate with CO2 near 400 ppm and warmer global temperatures, but paleoclimate data are sparse in current permafrost regions. Multiple proxies (carbonate δ18O, δ13C, CaCO3, TOC, TN, C/N, δ13Corg, grain size) and especially carbonate clumped isotopes (Δ47) enable quantitative paleotemperature reconstructions independent of water δ18O. Previous studies highlighted strong coupling of lake isotope signals to hydrology and productivity, and established relationships to estimate MAAT from lake temperatures. PlioMIP2 provides multi-model mPWP climate fields for contextualizing local reconstructions. Knowledge gaps include the stability and distribution of permafrost in past warm periods, and the relative vulnerability of alpine versus circumarctic permafrost carbon to warming.

Study site and sampling: The Kunlun Pass (KP) section on the northern Tibetan Plateau (35°39′N, 94°03′E, ~4.7 km elevation) comprises Plio–Pleistocene deep-lake deposits. The section spans 4.3–0.8 Ma per magnetostratigraphy and biostratigraphy. A total of 344 samples were analyzed for CaCO3 content, grain size, TOC, TN, C/N, and δ13Corg; 275 for bulk carbonate δ18O and δ13C; and 57 for clumped isotopes (Δ47). Petrography indicated dominantly micritic authigenic carbonates with limited diagenesis. Age model and cyclostratigraphy: Magnetic polarity intervals were correlated to GPTS and tied with biostratigraphy. Piecewise linear interpolation yielded the 4.3–0.8 Ma age model. Spectral analysis (Acycle) of CaCO3 and isotopes on the untuned series resolved eccentricity, obliquity, and precession signals, supporting age control. Geochemical measurements: Carbonate δ18O and δ13C were measured on CF-IRMS; analytical precision ±0.1‰ and ±0.06‰, respectively. Δ47 analyses followed carbonate-based standardization in the absolute reference frame using multiple internal standards, acid digestion at 90 °C, and “Easotope” processing; temperature was calculated with the Bernasconi–Müller calibration. To minimize detrital carbonate contamination, samples were sieved, and replicate δ18O–Δ47 temperature cross-plots identified authigenic end-members (higher δ18O, lower Δ47-T). Two outlier high-temperature samples were excluded. Proxy interpretations: Bulk carbonate δ18O and δ13C covary (r>0.7), indicating a closed-basin lake where E/P controls isotopic enrichment. CaCO3 reflects productivity and DIC. TOC, TN, and C/N distinguish organic matter sources (mostly algal; mean C/N ~5.6). δ13Corg informs macrophyte vs algal dominance and lake level changes. Grain size records erosional input linked to glaciation. Temperature reconstructions: Δ47-derived temperatures represent summer lake surface temperature (SLST). SLST was converted to local MAAT using a published transfer function (SLST–MAAT difference ~13–17 °C). Lake water δ18Ow was reconstructed from Δ47 temperature and measured carbonate δ18O. Statistical tests: Student’s t-tests across multiple potential boundaries (4.0–1.0 Ma) assessed step changes; p-values were smallest at 2.7 Ma across proxies. Bootstrap loess (1 Myr windows) summarized trends. Climate model integration: PlioMIP2 multi-model mean MAAT fields for mPWP (Eoi400) and pre-industrial (E280) were used to map −2 to 0 °C isotherms. The modern 0 °C MAAT isotherm best matches observed permafrost extent; thus 0 °C was adopted as a conservative threshold for permafrost persistence. Permafrost area and SOC calculations: Modern permafrost distribution (Obu et al.) was partitioned into middle-to-high latitude circumarctic (≥50°N) and low-to-middle latitude alpine regions (Tibetan Plateau–Pamir, Altai–Mongolia–Yablonoi–Sayan, Tian Shan, Rockies, Alps, Caucasus). Using ArcGIS, areas within the model 0 °C isotherm for mPWP-like and pre-industrial-like climates were computed and compared to modern to estimate destabilized permafrost areas. Potentially affected SOC was computed by multiplying thawed area by regional mean SOC densities: circumarctic SOC ~1300 Pg (65±6 kg m−2) and region-specific alpine SOC (e.g., Tibetan Plateau–Pamir 160±87 Pg), with uncertainties propagated via Monte Carlo. Elevation amplification assessment: A modern correlation between global temperature anomalies and northern Tibetan Plateau MAAT (ΔTNTP = 2.5725×ΔTG + 0.0295; r=0.84) was used to infer local amplification relative to estimated global Plio–Pleistocene cooling (~2–4 °C).

• At KP, multiple proxies show a marked shift at 2.7–2.6 Ma coincident with intensified Northern Hemisphere glaciation.
• Δ47 results indicate stepwise cooling: mean SLST ~15.0±6.9 °C (4.3–2.7 Ma) dropping to ~9.0±6.8 °C (2.7–0.8 Ma). Corresponding MAAT averaged 1.7 °C (4.3–2.7 Ma), decreased to −6.4 °C near 2.7–2.6 Ma, and averaged −7.7 °C thereafter, similar to modern local MAAT (−6 to −7 °C).
• Net local MAAT decrease of ~8.1 °C at 2.7–2.6 Ma (Δ47-based), consistent with global cooling and high-elevation amplification.
• Pliocene (pre-2.7 Ma) MAAT >0 °C at KP implies a permafrost-free northern Tibetan Plateau margin during Pliocene warmth; consistent with PlioMIP2 mPWP MAAT fields placing the 0 °C isotherm near the site.
• Under a mPWP-like climate: approximately 20% of circumarctic permafrost (~3.9×10^12 m^2) and ~60% of alpine permafrost (~1.9×10^12 m^2) would be destabilized.
• Potential SOC affected: circumarctic ~254±23 Pg; alpine ~85±60 Pg. Though alpine permafrost holds ~13–14% of total modern permafrost SOC, it could contribute ~25% of thaw-exposed carbon in a mPWP-like world.
• Lake water δ18Ow decreased from −4.3±1.8‰ (4.3–2.7 Ma) to −7.4±2.6‰ at 2.7–2.6 Ma, consistent with reduced evaporation during cooling.

The results directly address whether permafrost persisted in past warm climates and which regions are most vulnerable under future warming analogs. The KP reconstruction demonstrates Pliocene MAAT above freezing in the northern Tibetan Plateau, implying absent alpine permafrost during sustained warmth, followed by a rapid ~8 °C cooling at 2.7–2.6 Ma aligning with global glaciation intensification. Agreement between site-specific MAAT estimates and PlioMIP2 mPWP isotherms lends confidence to using the 0 °C MAAT threshold and model fields to project permafrost stability. Applying this framework globally indicates alpine permafrost is more thermally vulnerable than circumarctic permafrost in a mPWP-like climate, with ~60% of alpine permafrost area crossing the 0 °C threshold versus ~20% of circumarctic. Consequently, alpine regions, particularly the Tibetan Plateau, could disproportionately contribute to thaw-exposed SOC (~85 Pg), potentially representing ~25% of the total permafrost carbon exposed despite their smaller overall carbon stocks. The high-elevation amplification inferred from modern relationships supports larger temperature swings in alpine regions, explaining their outsized vulnerability. While the precise conversion of thawed SOC to atmospheric greenhouse gases depends on ecological and geomorphic factors, the spatial identification of vulnerable regions highlights where permafrost–carbon feedbacks may be most pronounced.

This study combines lacustrine clumped isotope thermometry with multi-proxy analyses and PlioMIP2 climate simulations to reconstruct Plio–Pleistocene temperatures on the Tibetan Plateau and evaluate permafrost stability under a mPWP-like climate. Key contributions include evidence for Pliocene MAAT >0 °C (permafrost-free) at the KP site, a ~8 °C cooling at 2.7–2.6 Ma tied to Northern Hemisphere glaciation, and a global assessment indicating that alpine permafrost could account for ~25% of thaw-exposed SOC (~85 Pg) in a warmer-than-present world despite comprising only ~13–14% of total permafrost SOC. The findings underscore the disproportionate vulnerability and potential climate feedback from alpine permafrost regions. Future research should: (1) expand paleotemperature reconstructions across permafrost regions to refine thresholds and regional variability; (2) improve coupling between permafrost area change and carbon release through process-based models capturing hydrology, abrupt vs gradual thaw, and ecosystem feedbacks; (3) quantify high-elevation amplification mechanisms; and (4) implement long-term monitoring in alpine regions to constrain SOC stocks, lateral export, and greenhouse gas fluxes under different warming scenarios.

• The 0 °C MAAT isotherm is a conservative, simplified threshold; permafrost persistence depends on vegetation, soil properties, moisture, snow cover, and ground conditions, leading to regional deviations of several degrees.
• Modern permafrost distribution reflects legacies of past glaciations, complicating analog-based inferences.
• Limited spatial coverage: primary paleoclimate constraints derive from a single Tibetan Plateau site; broader regional data would improve generalizability.
• Detrital carbonate contamination in lacustrine carbonates is mitigated but not entirely eliminable; Δ47 temperature calibration choice introduces small systematic shifts (conclusions robust across calibrations).
• The relationship between thawed SOC and greenhouse gas emissions is not well constrained and varies with topography, hydrology, and disturbance; estimates here reflect potentially affected SOC, not realized emissions or timing.
• Other destabilization processes (e.g., coastal erosion, infrastructure disturbance, landslides) are not explicitly modeled.
• Use of mPWP as a future analog captures first-order warmth but may differ in regional forcings and boundary conditions from anthropogenic warming trajectories.