Permafrost is ground that remains at or below 0°C for at least two consecutive years. It is not simply ice. It is a mixture of frozen soil, rock, ice, and organic matter that accumulated over millennia because cold temperatures slowed decay to a crawl. Today it underlies large parts of Siberia, Alaska, northern Canada, Greenland, high mountains in Europe and Asia, the Tibetan Plateau, and even pockets of Antarctica and the Andes. The broader permafrost region covers on the order of 15 to 20 million square kilometres. Locked inside it is an immense carbon store—commonly estimated at about 1,500 gigatonnes, roughly two to three times the carbon in all living vegetation on Earth.
That store is no longer securely frozen. Arctic air temperatures have risen at several times the global average. Heatwaves, heavier summer rainfall, longer thaw seasons, and more frequent wildfires are deepening the active layer—the surface zone that thaws in summer and refreezes in winter—and warming the ground below. A 2026 global assessment of 156 long-term monitoring sites found statistically significant thickening of the active layer at 55 percent of Arctic sites and 38 percent of Antarctic sites between 2000 and 2024. In European mountain permafrost and high-elevation Asian sites, more than 90 percent of records showed thickening. Average Arctic rates were about 0.8 centimetres per year; European mountain sites averaged more than 10 centimetres per year, and some Swiss Alpine boreholes recorded extreme deepening. Southern-margin permafrost in Fennoscandia is already committed to loss even if temperatures were stabilized today.
Two styles of thaw matter. Gradual thaw lowers the permafrost table from the top down as summers grow warmer and wetter. Abrupt thaw is faster and more destructive. In ice-rich ground, melting of massive ice wedges and lenses causes the surface to collapse into thermokarst lakes, troughs, and retrogressive thaw slumps. Those features expose deeper, older carbon, alter drainage, and can persist as local sources of greenhouse gases for decades. Wildfire compounds both pathways. Fire removes insulating vegetation and organic soil, darkens the surface, and can trigger or accelerate thaw. Models that couple permafrost, hydrology, and atmosphere suggest that rapid thaw can dry soils, raise surface temperatures, and produce abrupt increases in burned area across parts of the Arctic–Subarctic. Fire carbon release and post-fire thaw together amplify the climate feedback beyond gradual thaw alone.
The climate concern is the permafrost carbon feedback. Once ice melts, microbes consume previously frozen organic matter and respire carbon dioxide under aerobic conditions or methane under waterlogged ones. Methane is far more potent than carbon dioxide over decades, so even a small share of carbon leaving as methane can dominate near-term radiative forcing. IPCC assessments concluded with high confidence that thawing terrestrial permafrost will release carbon, but with low confidence in exact timing, magnitude, and the split between CO₂ and CH₄. A widely cited synthesis puts the gradual-thaw feedback near 18 petagrams of carbon per degree of global warming by 2100, with a wide range. Including abrupt thaw and wildfire can raise that number substantially and shrink remaining carbon budgets consistent with 1.5°C or 2°C. One recent modelling exercise found that adding those processes reduced remaining allowable budgets from 2025 onward by roughly a quarter for 1.5°C and about 17 percent for 2°C. Other work suggests that under high-emission pathways the northern soil carbon balance could flip from sink to source earlier in this century than older models implied.
The feedback is real and positive, but it is not a single global on–off switch. A 2024 perspective in Nature Climate Change argued that despite rapid, locally irreversible ice loss and thermokarst, the pan-Arctic response remains roughly linear with warming. There is no safe margin in which further warming would leave permafrost carbon untouched. Each additional fraction of a degree exposes more ground. Local tipping behaviour—collapse of ice-rich slopes, drainage of lakes, sudden methane bursts from new ponds—does not add up to a self-sustaining planetary runaway on decadal timescales. That distinction is important. It means the problem cannot be dismissed as either “already lost” or “only a future cliff.” It is a continuous tax on the remaining carbon budget.
Impacts are already visible beyond the atmosphere. In northern Alaska, rivers are carrying more dissolved organic carbon as thaw seasons stretch into autumn; some of that carbon is converted to CO₂ in the Arctic Ocean. Siberian methane emissions during the thaw season have risen in recent analyses. On the Tibetan Plateau, ground-ice melt contributes to lake expansion and alters runoff in major Asian river headwaters, while thaw slumps can flip alpine grassland patches from carbon sinks toward sources. Coastal erosion of ice-rich bluffs is expected to accelerate, delivering more carbon to the sea and reducing the Arctic Ocean’s capacity to take up atmospheric CO₂.
Infrastructure is the most immediate human cost. Buildings, roads, pipelines, and airstrips in the Arctic were designed for frozen ground that does not settle. As ice melts, foundations shift, roads buckle, and pipelines lose alignment. Studies have long warned that a large share of Arctic infrastructure could face high risk by mid-century; community-scale work now documents places where damage is arriving decades earlier than regional models predicted. Point Lay, Alaska, is one example: ice-wedge thermokarst now underlies most of the residential area, damaging housing and complicating subsistence access. Along the Qinghai–Tibet engineering corridor, high-resolution projections under high emissions show sharp expansion of high-risk zones and multi-billion-dollar additional replacement costs later this century. Transportation networks between roughly 40°N and 65°N are especially exposed.
There are further, less quantified risks. Frozen ground has sequestered not only carbon but also contaminants, ancient microbes, and, in some locations, industrial and radioactive residues. Thaw can remobilize mercury and other pollutants into food webs. Arctic ecosystems that evolved with a reliable frozen substrate face hydrologic rearrangement: some landscapes wet, others drain and dry; plant communities shift from tundra toward shrub and, in places, boreal forest. Those vegetation changes can either shade and insulate remaining permafrost or darken the surface and accelerate thaw, depending on local conditions.
What follows from this is straightforward even if the numbers remain uncertain. Permafrost thaw is already underway on every continent that has permafrost. Mountain sites are changing fastest in the observational record; Arctic lowlands hold the largest carbon inventories. Emissions from thaw will not dwarf fossil-fuel emissions in this century, but they are large enough to matter for Paris-aligned budgets, and they continue after human emissions fall because of thermal inertia in the ground. Some southern and warm permafrost is already committed. Reducing global warming still limits how much additional ground thaws, how much abrupt collapse occurs, and how much fire and hydrology amplify the release. Monitoring networks, high-resolution mapping of ice-rich terrain, and models that include thermokarst and fire are improving the picture. The physical process itself will not wait for perfect accounting. Frozen carbon, once thawed and decomposed, does not refreeze on human timescales.