Positive feedbacks between permafrost degradation and the release of soil carbon into the atmosphere impact land-atmosphere interactions, disrupt the global carbon cycle, and accelerate climate change. The widespread distribution of thawing permafrost is causing a cascade of geophysical and biochemical disturbances with global impacts. Currently, few earth system models account for permafrost carbon feedback (PCF) mechanisms. This research study integrates artificial intelligence (AI) tools and information derived from field-scale surveys across the tundra and boreal landscapes in Alaska. We identify and interpret the permafrost carbon cycling links and feedback sensitivities with GeoCryoAI, a hybridized multimodal deep learning (DL) architecture of stacked convolutionally layered, memory-encoded recurrent neural networks (NN). This framework integrates in-situ measurements and flux tower observations for teacher forcing and model training. Preliminary experiments to quantify, validate, and forecast permafrost degradation and carbon efflux across Alaska demonstrate the fidelity of this data-driven architecture. More specifically, GeoCryoAI logs the ecological memory and effectively learns covariate dynamics while demonstrating an aptitude to simulate and forecast PCF dynamics-active layer thickness (ALT), carbon dioxide flux (CO2), and methane flux (CH4)-with high precision and minimal loss (i.e. ALTRMSE: 1.327 cm [1969-2022]; CO2 RMSE: 0.697 mu molCO2m-2s-1 [2003-2021]; CH4 RMSE: 0.715 nmolCH4m-2s-1 [2011-2022]). ALT variability is a sensitive harbinger of change, a unique signal characterizing the PCF, and our model is the first characterization of these dynamics across space and time.

A significant portion of the large amount of carbon (C) currently stored in soils of the permafrost region in the Northern Hemisphere has the potential to be emitted as the greenhouse gases CO2 and CH4 under a warmer climate. In this study we evaluated the variability in the sensitivity of permafrost and C in recent decades among land surface model simulations over the permafrost region between 1960 and 2009. The 15 model simulations all predict a loss of near-surface permafrost (within 3m) area over the region, but there are large differences in the magnitude of the simulated rates of loss among the models (0.2 to 58.8x10(3)km(2)yr(-1)). Sensitivity simulations indicated that changes in air temperature largely explained changes in permafrost area, although interactions among changes in other environmental variables also played a role. All of the models indicate that both vegetation and soil C storage together have increased by 156 to 954TgCyr(-1) between 1960 and 2009 over the permafrost region even though model analyses indicate that warming alone would decrease soil C storage. Increases in gross primary production (GPP) largely explain the simulated increases in vegetation and soil C. The sensitivity of GPP to increases in atmospheric CO2 was the dominant cause of increases in GPP across the models, but comparison of simulated GPP trends across the 1982-2009 period with that of a global GPP data set indicates that all of the models overestimate the trend in GPP. Disturbance also appears to be an important factor affecting C storage, as models that consider disturbance had lower increases in C storage than models that did not consider disturbance. To improve the modeling of C in the permafrost region, there is the need for the modeling community to standardize structural representation of permafrost and carbon dynamics among models that are used to evaluate the permafrost C feedback and for the modeling and observational communities to jointly develop data sets and methodologies to more effectively benchmark models.

Release of greenhouse gases from thawing permafrost is potentially the largest terrestrial feedback to climate change and one of the most likely to occur; however, estimates of its strength vary by a factor of thirty. Some of this uncertainty stems from abrupt thaw processes known as thermokarst (permafrost collapse due to ground ice melt), which alter controls on carbon and nitrogen cycling and expose organic matter from meters below the surface. Thermokarst may affect 20-50% of tundra uplands by the end of the century; however, little is known about the effect of different thermokarst morphologies on carbon and nitrogen release. We measured soil organic matter displacement, ecosystem respiration, and soil gas concentrations at 26 upland thermokarst features on the North Slope of Alaska. Features included the three most common upland thermokarst morphologies: active-layer detachment slides, thermo-erosion gullies, and retrogressive thaw slumps. We found that thermokarst morphology interacted with landscape parameters to determine both the initial displacement of organic matter and subsequent carbon and nitrogen cycling. The large proportion of ecosystem carbon exported off-site by slumps and slides resulted in decreased ecosystem respiration postfailure, while gullies removed a smaller portion of ecosystem carbon but strongly increased respiration and N2O concentration. Elevated N2O in gully soils persisted through most of the growing season, indicating sustained nitrification and denitrification in disturbed soils, representing a potential noncarbon permafrost climate feedback. While upland thermokarst formation did not substantially alter redox conditions within features, it redistributed organic matter into both oxic and anoxic environments. Across morphologies, residual organic matter cover, and predisturbance respiration explained 83% of the variation in respiration response. Consistent differences between upland thermokarst types may contribute to the incorporation of this nonlinear process into projections of carbon and nitrogen release from degrading permafrost.

Degrading permafrost can alter ecosystems, damage infrastructure, and release enough carbon dioxide (CO2) and methane (CH4) to influence global climate. The permafrost carbon feedback (PCF) is the amplification of surface warming due to CO2 and CH4 emissions from thawing permafrost. An analysis of available estimates PCF strength and timing indicate 120 +/- 85 Gt of carbon emissions from thawing permafrost by 2100. This is equivalent to 5.7 +/- 4.0% of total anthropogenic emissions for the Intergovernmental Panel on Climate Change (IPCC) representative concentration pathway (RCP) 8.5 scenario and would increase global temperatures by 0.29 +/- 0.21 degrees C or 7.8 +/- 5.7%. For RCP4.5, the scenario closest to the 2 degrees C warming target for the climate change treaty, the range of cumulative emissions in 2100 from thawing permafrost decreases to between 27 and 100 Gt C with temperature increases between 0.05 and 0.15 degrees C, but the relative fraction of permafrost to total emissions increases to between 3% and 11%. Any substantial warming results in a committed, long-term carbon release from thawing permafrost with 60% of emissions occurring after 2100, indicating that not accounting for permafrost emissions risks overshooting the 2 degrees C warming target. Climate projections in the IPCC Fifth Assessment Report (AR5), and any emissions targets based on those projections, do not adequately account for emissions from thawing permafrost and the effects of the PCF on global climate. We recommend the IPCC commission a special assessment focusing on the PCF and its impact on global climate to supplement the AR5 in support of treaty negotiation.