Significant progress in permafrost carbon science made over the past decades include the identification of vast permafrost carbon stocks, the development of new pan‐Arctic permafrost maps, an increase in terrestrial measurement sites for CO₂ and methane fluxes, and important factors affecting carbon cycling, including vegetation changes, periods of soil freezing and thawing, wildfire, and other disturbance events. Process‐based modeling studies now include key elements of permafrost carbon cycling and advances in statistical modeling and inverse modeling enhance understanding of permafrost region C budgets. By combining existing data syntheses and model outputs, the permafrost region is likely a wetland methane source and small terrestrial ecosystem CO₂ sink with lower net CO₂ uptake toward higher latitudes, excluding wildfire emissions. For 2002–2014, the strongest CO₂ sink was located in western Canada (median: −52 g C m⁻² y⁻¹) and smallest sinks in Alaska, Canadian tundra, and Siberian tundra (medians: −5 to −9 g C m⁻² y⁻¹). Eurasian regions had the largest median wetland methane fluxes (16–18 g CH₄ m⁻² y⁻¹). Quantifying the regional scale carbon balance remains challenging because of high spatial and temporal variability and relatively low density of observations. More accurate permafrost region carbon fluxes require: (a) the development of better maps characterizing wetlands and dynamics of vegetation and disturbances, including abrupt permafrost thaw; (b) the establishment of new year‐round CO₂ and methane flux sites in underrepresented areas; and (c) improved models that better represent important permafrost carbon cycle dynamics, including non‐growing season emissions and disturbance effects.

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. ALT^RMSE: 1.327 cm [1969–2022]; CO₂ ^RMSE: 0.697 µmolCO₂m⁻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.

Permafrost degradation in the Arctic is accelerating and is forecast to enhance greenhouse gas (GHG) emissions from the large permafrost carbon pool. Earth Observation has a key role in determining GHG sources and sinks, and multiple current and future missions are useful to track baseline parameters for determining GHG fluxes. NASA and ESA have initialized the Arctic Methane and Permafrost Challenge (AMPAC) as a transatlantic networking action that strives to promote related scientific work and improve observation capabilities. Key variables observable from space include methane concentrations as well as landcover properties to inform process-based models as proxy for sources as well as temperature-related constraints for microbial activity. Upcoming missions are expected to advance these capabilities significantly with increased sampling intervals through future synthetic aperture radar missions and constellations of multiple multispectral sensors. This will allow better representation of seasonality and advance methane source attribution in general. In addition, continuity of current missions which provide greenhouse gas observations, including methane, is crucial. Hyper- and superspectral sensors targeting primarily landsurface observation are expected to complement methane retrievals through identification of emission hotspots. Arctic monitoring also requires active optical instruments for concentration retrieval, a type of instrumentation which is still novel for satellite-based observations. A comprehensive portfolio of hyper-spectral, passive microwave, synthetic aperture radar, altimeter and landsurface temperature and lidar measurements in addition to imaging spectrometers will be available by 2032/33, at the time of the International Polar Year. This will allow for advanced experiments when also accompanying in situ observations become available.

Given a world increasingly dominated by climate extremes, modifying the Earth's climate with large-scale geoengineering intervention is inevitable. However, geoengineering faces a conundrum: forecasting the consequences of climate intervention accurately in a system for which we have incomplete observations and an imperfect understanding. We evaluate the global response and potential implications of mitigation and intervention deployment by utilizing CRU TS4.08 observations, ERA5 reanalysis data, and CMIP6 scenario-based UKESM0-1-LL simulations. From 1950 to 2022, global weighted mean surface temperature (Tsurf) and total precipitation (P) rose by 1.37[Formula: see text]0.48 °C and 0.05[Formula: see text]0.57 mm day-1. Significant regional Tsurf anomalies and erratic interannual variability of P were revealed, with ranges from 7.63 °C in Greenland and northern Siberia to -2.38 °C in central Africa and 1.17 mm day-1 in southern Alaska to -1.20 mm day-1 in Colombia and east Africa. Collectively, mitigation and intervention simulations tended to overestimate the variability and magnitude of Tsurf and P, exhibiting substantial regional discrepancies and scenario-specific heterogeneity when estimating atmospheric methane concentration ([CH₄]). Despite capturing significant departures in Tsurf, P, and [CH₄], replicating historical P teleconnections and spatial patterns of warming remained a challenge. These results underscore regional disparities with global implications, harkening the necessity to refine existing architectures while developing novel methods to evaluate the risks and feasibility of geoengineering intervention.

Complex nonlinear relationships exist between the permafrost thermal state, active layer thickness, and terrestrial carbon cycle dynamics. In Arctic and boreal Alaska, significant uncertainties characterize the spatiotemporal rate and magnitude of permafrost degradation and the permafrost carbon feedback, with increasing recognition of the importance of thawing mechanisms. The challenges of monitoring sub‐surface phenomena with remote sensing technology further complicate the issue. There is an urgent need to understand how and to what extent thawing permafrost destabilizes the carbon balance in Alaska and to characterize the feedback involved. In this research, we use our artificial intelligence‐driven model GeoCryoAI to quantify permafrost carbon dynamics in Alaska. The GeoCryoAI model uses a hybridized process‐constrained ensemble learning framework to simultaneously ingest, scale, and analyze in situ measurements, remote sensing observations, and process‐based modeling outputs with disparate spatiotemporal sampling and data densities. We evaluated prior naïve (a) persistence and (b) teacher forcing approaches relative to (c) time‐delayed GeoCryoAI simulations, yielding the following error metrics (RMSE) for active layer thickness (ALT), methane (CH₄), and carbon dioxide (CO₂), respectively: 1.997, 1.327, 1.007 cm [1963–2022]; 0.884, 0.715, 0.694 nmol CH₄km⁻² month⁻¹ [1994–2022]; 1.906, 0.697, 0.213 µmol CO₂km⁻² month⁻¹ [1994–2022]. Our approach overcomes traditional model inefficiencies and resolves spatiotemporal disparities. GeoCryoAI captures abrupt and persistent changes while introducing a novel methodology for assimilating contemporaneous information at various scales. We describe GeoCryoAI, the methodology, our results, and plans for future applications.