An earth system model shows self-sustained thawing of permafrost even if all man-made GHG emissions stop in 2020

The paper examines whether the Earth’s climate system has already crossed a point-of-no-return leading to self-sustained warming, with a focus on permafrost thaw. Points-of-no-return are thresholds that, once crossed, can trigger irreversible processes such as permafrost thawing, forest dieback, or ocean acidification. Using the reduced-complexity earth system model ESCIMO, run from 1850 to 2500, the authors investigate whether global temperature can continue to rise even after anthropogenic greenhouse gas (GHG) emissions cease. They report that in ESCIMO the global temperature keeps increasing for centuries regardless of how quickly humanity reduces GHG emissions, driven by a reinforcing cycle involving declining surface albedo (from melting ice and snow), increasing atmospheric water vapour (from higher temperatures), and carbon releases (CH4 and CO2) from thawing permafrost. This self-sustaining cycle appears to be triggered by as little as +0.5 °C warming above pre-industrial levels.

The study builds on decades of discussion regarding climate tipping points and points-of-no-return, referencing syntheses that warn such thresholds may be nearer than assumed (e.g., Lenton et al., 2019). It situates ESCIMO’s findings within this context and compares model outputs with other permafrost-carbon models summarized by McGuire et al. (2018). The comparison indicates ESCIMO tends to produce larger cumulative carbon releases from permafrost by 2300 than many other models, though differences in forcing scenarios (Scenario 1 vs. RCP4.5) complicate direct comparison. Prior foundational work on radiative forcing and water vapour feedbacks (Hansen, Rind, Ramanathan and Inamdar, Cess) informs the model’s treatment of feedback processes.

The authors use ESCIMO, a reduced-complexity, system-dynamics Earth system model representing the atmosphere, oceans, land types (including forests and tundra), biomass, permafrost, and their interactions. Simulations run from 1850 to 2500 under two emissions scenarios: Scenario 1 assumes anthropogenic GHG emissions peak in the 2030s and decline to zero by 2100; Scenario 2 assumes emissions are cut to zero in 2020. After reaching zero, emissions remain zero. Key processes include GHG-driven radiative forcing, water vapour feedback, sea-level rise from thermal expansion and glacier melt, and surface albedo changes from ice and snow loss. Permafrost thaw is modeled as heat transfer from atmosphere to frozen soil, with thaw rate proportional to the temperature difference between air and frozen soil; post-thaw tundra begins absorbing CO2 via photosynthesis with rates dependent on temperature and atmospheric CO2. ESCIMO estimates the relative contributions to energy balance by inferring radiative effects: for well-mixed GHGs (CO2, CH4, N2O, other gases) using IPCC formulas, and for feedbacks (water vapour and albedo) via proxies based on changes in top-of-atmosphere longwave radiation and surface shortwave reflection relative to 1850. Surface albedo forcing is estimated by comparing shortwave reflection over time to the 1850 baseline, accounting for land and ocean albedo (ocean albedo declines as Arctic ice melts). Sensitivity analyses: (1) Latin-Hypercube sampling of 14 uncertain parameters with ±10% uniform ranges, 200 runs, to test robustness of temperature trajectories; (2) targeted parameter sweeps central to permafrost thaw: (a) fraction of carbon released as CH4 vs CO2 from permafrost (entire range 0–100%, and restricted 0–15% informed by literature); (b) slope of permafrost thaw rate vs temperature, with a reference full-depth thaw rate of 12,500 km²/yr at 4 °C and a linear temperature scaling; (c) slope of the future relationship between added water vapour and additional blocking of outgoing longwave radiation beyond observed humidity, parameterized by a third-order polynomial calibrated to historical global averages. Additional exploratory experiments tested counterfactual timing of emission cessation (to identify thresholds) and evaluated the scale of annual CO2 removal required from 2020 onward to halt self-sustained warming.

• In both scenarios, global mean temperature continues rising for centuries after anthropogenic emissions cease, reaching about +3 °C by 2500; sea level rises monotonically to roughly +3 m by 2500.
• Scenario 1: Temperature peaks around 2075 at +2.3 °C, then declines to +2.0 °C by ~2150 as atmospheric GHG concentrations fall and energy is used to melt ice. After ~2150, temperatures resume rising despite continued zero emissions and declining atmospheric CO2 due to a self-reinforcing cycle: decreasing surface albedo, increasing water vapour, and carbon release (CH4 and CO2) from thawing permafrost.
• Scenario 2: Even with emissions cut to zero in 2020, temperatures again rise after ~2150, showing the same self-sustained warming cycle.
• Albedo changes: In ESCIMO, average ocean albedo declines from 0.080 (2070) to 0.067 (2300), and surface albedo from 0.127 to 0.117, increasing absorbed shortwave energy by about 1.7 W/m² (radiative effect of delta albedo rises from ~0.8 W/m² in 2070 to ~2.6 W/m² in 2300).
• Water vapour feedback remains strong and does not vanish when atmospheric CO2 returns toward pre-industrial levels, so long as temperatures remain elevated; after ~2150, water vapour’s contribution to trapping is comparable to the sum of other GHGs.
• Permafrost carbon: ESCIMO Scenario 1 yields cumulative permafrost carbon release of ~175 GtC by 2300 and thawing of ~2 million km², which is on the higher end relative to other models (3–5 million km² thaw area; 66±70 GtC release in other studies, with varying vegetation uptake).
• Sensitivity analyses (200 runs, ±10% on 14 parameters) show absolute temperature levels vary, but the qualitative behavior—self-sustained post-2150 warming—persists in 75% of runs. Targeted parameter sweeps (fraction of CH4 release, thaw-rate slope, water vapour blocking slope) significantly affect rates but do not remove the self-sustained warming pattern.
• Counterfactual analysis suggests avoiding the self-reinforcing permafrost-thaw cycle would have required anthropogenic emissions to fall to zero between 1960 and 1970 (when warming was <~0.5 °C above pre-industrial).
• Mitigation via CO2 removal: At least ~33 GtCO2e per year of sustained removal from 2020 onward (e.g., large-scale direct air capture or bioenergy with CCS) would be required in ESCIMO to stop the self-sustained temperature rise; this is deemed technically feasible but extremely costly and with potential side effects for some geoengineering options.

The findings suggest that a point-of-no-return has been crossed in the ESCIMO framework: once initiated by modest early warming (~+0.5 °C), coupled feedbacks—declining albedo, increased atmospheric water vapour, and permafrost carbon releases—maintain and amplify warming independent of ongoing anthropogenic emissions. The continued warming after atmospheric CO2 declines underscores that water vapour and albedo feedbacks can sustain elevated temperatures, with water vapour’s radiative effect comparable to all other GHGs combined after ~2150. The drop in surface and ocean albedo increases shortwave absorption sufficiently to re-accelerate warming after the mid-22nd century. Comparison with other models indicates ESCIMO may produce higher permafrost carbon releases under its scenario, though scenario differences and aggregation limit direct comparability. The significance lies in highlighting a potential long-term commitment to warming and permafrost thaw even under aggressive emissions cuts, emphasizing the need to consider carbon removal and non-CO2 feedbacks in policy and modeling.

Self-sustained thawing of permafrost is a robust outcome in the ESCIMO model under a wide range of plausible parameter values and emissions scenarios, disappearing only under counterfactual early cessation of emissions (1960s) or under parameterizations that fail to reproduce historical climate. The study contributes evidence, from a reduced-complexity model, that coupled feedbacks could commit the Earth system to centuries of additional warming and sea-level rise even after emissions stop. The authors call for other modeling groups using larger and more detailed models to test and report whether similar self-sustained dynamics emerge, and to explore intervention strategies, including the feasibility, scale, and side effects of carbon removal and geoengineering.

ESCIMO is a reduced-complexity, highly aggregated system-dynamics model that simplifies many processes (e.g., permafrost thaw mechanisms, regional heterogeneity, cloud processes) and is not sufficiently regionalized to resolve spatial uptake and release dynamics. Some parameterizations (e.g., historical assumption of all permafrost carbon released as CH4 in the base case; simplified thaw-rate linear scaling; polynomial extrapolation for water vapour blocking beyond observed humidity) introduce uncertainties. Absolute projections (e.g., magnitude of temperature and carbon release) are sensitive to parameter choices, and differences in forcing scenarios hinder direct comparison with other models. While parameter sets that remove the self-sustained warming can be devised, they tend not to reproduce the historical period accurately in ESCIMO.