Timescales of the permafrost carbon cycle and legacy effects of temperature overshoot scenarios

Permafrost regions in the high northern latitudes store vast organic carbon pools protected by low temperatures. Arctic warming reduces permafrost extent and thickness, exposing more soil organic matter (SOM) to decomposition and creating a positive feedback via increased CO2 and CH4 emissions. Although many studies have assessed permafrost degradation under warming scenarios, substantial uncertainty remains about long-term responses and how quickly soils and carbon fluxes adapt to stabilized climates, given long carbon turnover times and strong thermophysical inertia. A crucial open question is whether the eventual steady-state depends on the prior climate trajectory, especially under temperature overshoot (OS) pathways that temporarily exceed targets before stabilizing at 1.5 °C (PACT1.5). Prior modeling often neglected SOM effects on soil thermal properties. This study investigates (1) the timescales of adjustment of high-latitude ecosystems and permafrost soils to PACT1.5 and (2) whether OS scenarios leave irreversible legacy effects and allow multiple steady-states via feedbacks among water, energy, and carbon cycles.

The paper synthesizes evidence that northern permafrost regions hold around 1100–1700 GtC and are warming twice as fast as the global mean, elevating concern about permafrost as a climate tipping element. Previous studies examined SOM loss and GHG release under varied warming scenarios, but the long-term adaptation to stable climates remains uncertain, particularly where cold soils slow turnover and high water contents increase thermal inertia. Some prior work suggested path independence and only transient hysteresis in permafrost equilibrium states, and indicated convergence of global carbon budgets across OS scenarios. However, these studies typically lacked explicit feedbacks linking SOM to soil thermal and hydrological properties, which can influence carbon accumulation rates. The present work addresses this gap by incorporating such feedbacks.

Model: JSBACH (land surface component of MPI-ESM1.2) in an offline (uncoupled) setup with prescribed atmospheric forcing, substantially adapted for high-latitude processes. Key adaptations include: vertically discretized soil carbon pools (YASSO-based decomposition with temperature and moisture dependence using soil, not surface, temperature; simplified CENTURY-type moisture limitation); separation of inundated vs. non-inundated fractions for anoxic/oxic decomposition; vertical SOM transport via cryo/bio-turbation as diffusion dependent on active layer saturation and freeze–thaw days; modified soil physics following Ekici et al. with multi-layer snow, phase change of water, and water effects on thermal properties; explicit SOM effects on thermal/hydrological properties; allowance for supercooled liquid water; revised infiltration from temperature- to saturation-based formulation; water stress computed relative to the fraction of root zone above the permafrost table; implementation of two wetland formation schemes (surface ponding and TOPMODEL-based saturation). No peatland-specific vegetation parametrizations were included. Initialization and data: Soil C initialized using WISE30sec (0–2 m) and NCSCDv2 (2–3 m). To reduce underestimation of top 3 m soil C at coarse resolution, soil depth was extended (per Schneider von Deimling et al.) and above-bedrock pools were upscaled, yielding about 931 GtC in the upper 3 m (vs. 1015 GtC in combined datasets). Below 3 m initialized without organic matter. Initial simulation (NoOS) started in 1850 with observation-based C pools and preindustrial physical state; ran through historical and SSP5-8.5 to 2035 to establish consistent active layer C and retain frozen C below. Forcing and resolution: Forcing from CMIP6 historical (1950–2014) and SSP5-8.5 (2015–2100). Stable PACT1.5 atmospheric conditions represented by cycling years 2030–2040 from MPI-ESM1.2. Horizontal grid T63 (~1.9°×1.9°); 18 subsurface layers to 100 m (11 layers in top 3 m); timestep 1800 s; spatial domain predominantly 45–90°N, with selected grid-cell extensions for very long runs. Scenarios: Four pathways to PACT1.5 for permafrost regions: (i) NoOS (reach PACT1.5 without prior overshoot), and OS peaks at 2050 (P2050), 2075 (P2075), and 2100 (P2100), all based on SSP5-8.5 then reversed to PACT1.5. After each, long constant-climate (CC) runs at PACT1.5 were conducted for up to 1000 years; additional runs isolated OS indirect effects by combining P2100 carbon pools with NoOS physical initial state, extended to 2500 years. Selected grid boxes were extended a further 7500 years (total 10,000 years) to test steady-state dependence on initial SOM. Adjustment timescale metric: Defined as time until the 100-year-running-mean rate of change declines by an order of magnitude relative to the first 100 years of the NoOS CC run, allowing cross-scenario comparison. Caveats tracked: Model lacks pedogenesis, erosion, dust deposition, and fixed soil depths over millennia; no coupled land–atmosphere feedbacks; coarse resolution; lacks small-scale geomorphological processes (e.g., thermokarst, subsidence).

• Slow adjustment: High-latitude ecosystems require centuries to adjust to stabilized PACT1.5 conditions. Even after 1000 years, minor trends persist in soils and fluxes.
• State at stabilization depends on pathway: After reaching PACT1.5, soil temperatures at 3 m differ by about 0.3–0.5 °C across pathways (10–30% of the overshoot warming). Total soil water differs by up to ~5 cm. SOM stocks differ by up to ~70 GtC; total terrestrial carbon differs by up to ~40 GtC.
• Net carbon fluxes: Without overshoot (NoOS), the permafrost region is a CO2 source at stabilization (~0.2 GtC yr−1). With prior OS, the region becomes a sink (up to ~0.3 GtC yr−1). Table values: NEF at PACT1.5: NoOS ~+0.18 GtC yr−1; P2050 ~+0.03; P2075 ~−0.26; P2100 ~−0.24.
• Adjustment timescales: Under NoOS, net ecosystem flux largely adjusts in ~450 years; soil temperatures at 3 m take ~400 years to adapt. Following a large OS (peak 2100), soil temperatures adjust within ~100 years. Soil water content adapts over several centuries in all cases and retains a drying trend.
• Dependence on soil type: Long trends occur mainly in organic-rich soils; mineral-dominated soils adjust rapidly to PACT1.5.
• Multistability and legacy effects: In many permafrost-retaining grid cells, differing initial SOM (pre- vs post-OS) leads to persistent differences in belowground temperatures, moisture, and SOM concentrations under identical PACT1.5 forcing, indicating multiple steady-states maintained by feedbacks among SOM insulation, hydrology, evapotranspiration, and nutrient cycling. Cooler subsurface can promote ongoing SOM accumulation; warmer subsurface can stabilize at lower SOM, even with continuous SOM loss in extremes.
• Direction of legacy effects: OS-induced SOM loss yields higher belowground temperatures (average +0.1 to +0.2 °C at 1 m in MJJASO), smaller near-surface permafrost volume, slightly lower total soil water, and higher soil respiration. Higher mineral N availability from enhanced decomposition partially alleviates nutrient limitation, increasing NPP; net effect is larger in- and outgoing soil C fluxes but lower equilibrium SOM stocks and lower total terrestrial C (~−40 GtC) relative to NoOS.
• Spatial variability: Legacy effects are strongest in organic-rich northern regions and North America; differences in SOM density >20 kg C m−2 after 2500 years are common in these areas. While the dominant effect is deeper active layers and warmer soils after OS, some areas show the opposite due to local hydrological states.
• Methane: Average wetland area changes little, but soil methane production declines by ~10% after OS due to altered overlap between inundation and high-SOM zones.
• Proportionality: Legacy effect magnitude scales with OS magnitude; even small OS can alter long-term steady-states in the Arctic.

The study demonstrates that high-latitude ecosystems and permafrost soils respond on multi-century timescales to stabilized 1.5 °C conditions, addressing the key question about adjustment speed and path dependence. The pathway to stabilization materially affects the state at stabilization and the ensuing long-term equilibrium, through OS-induced modifications of SOM that alter soil thermal and hydrological properties. These biogeophysical–biogeochemical feedbacks can maintain multiple steady-states, rendering OS legacy effects effectively irreversible under steady atmospheric conditions. Consequently, emission pathways that overshoot targets influence regional CO2 and CH4 balances for millennia, with implications for allowable carbon budgets and the design of mitigation pathways. The findings challenge assumptions of path independence for permafrost systems and highlight the need to consider permafrost feedbacks explicitly in policy-relevant carbon budget estimates.

This work shows that permafrost-affected ecosystems require centuries to millennia to equilibrate after climate stabilization at 1.5 °C, and that temperature overshoots leave enduring, effectively irreversible legacy effects mediated by SOM–thermo–hydro–carbon feedbacks. OS scenarios can shift regions into alternative steady-states with warmer subsurface, altered hydrology, higher respiration and productivity, and lower SOM and total terrestrial carbon, while reducing methane production by about 10% on average. These results imply that even modest overshoots can have long-term consequences for Arctic carbon–climate feedbacks and should be minimized in mitigation planning. Future research should: (i) incorporate land–atmosphere coupling to capture feedbacks (e.g., Bowen ratio effects on precipitation); (ii) resolve or parameterize small-scale hydrological and geomorphological processes (thermokarst, subsidence, thaw lakes) and peatland dynamics; (iii) perform multi-model and parameter-ensemble assessments to quantify uncertainty; and (iv) improve representation of pedogenesis, erosion, and evolving soil properties over millennia.

• Offline, uncoupled land model setup excludes land–atmosphere feedbacks that could amplify or dampen legacy effects (e.g., moisture–precipitation interactions).
• Coarse spatial resolution and absence of small-scale hydrological and geomorphological processes (thermokarst, ice-wedge degradation, ground subsidence, thaw lake dynamics) may miss abrupt, nonlinear changes and their reversibility.
• Single-model study without parameter ensemble; some processes (e.g., methane) are sensitive to parameter choices; broader model disagreement exists on future Arctic moisture trends.
• Long-timescale constraints: No pedogenesis, erosion, dust deposition; fixed soil depths; mineral soil properties held constant over millennia; below-3 m soil initialized without organic carbon (despite known deep stocks), potentially underestimating total C and responses.
• Simplified wetland representation and lack of peatland-specific vegetation parameters; vertical drainage and lateral flow approximations; rooting dynamics simplified via fixed depth and stress computed relative to permafrost table.
• Adjustment timescale definition based on a specific threshold (10× reduction in rate of change) may affect quantitative estimates.