SERGEY ZIMOV, AN ecologist by training, tossed a woolly mammoth bone on the pile. He was squatting in mud along the cool, wide Kolyma River, below a towering cliff of crumbling earth. It was summer in eastern Siberia, far above the Arctic Circle, in that part of Russia that’s closer to Alaska than to Moscow. There wasn’t a speck of frost or snow in sight. Yet at this cliff, called Duvanny Yar, the Kolyma had chewed through and exposed what lies beneath: a layer of frozen ground, or permafrost, that is hundreds of feet deep—and warming fast.

Twigs, other plant matter, and Ice Age animal parts—bison jaws, horse femurs, mammoth bones—spilled onto a beach that sucked at Zimov’s boots. “I love Duvanny Yar,” he said as he yanked fossils from the muck. “It is like a book. Each page is a story about the history of nature.”

Across nine million square miles at the top of the planet, climate change is writing a new chapter. Arctic permafrost isn’t thawing gradually, as scientists once predicted. Geologically speaking, it’s thawing almost overnight. As soils like the ones at Duvanny Yar soften and slump, they’re releasing vestiges of ancient life—and masses of carbon—that have been locked in frozen dirt for millennia. Entering the atmosphere as methane or carbon dioxide, the carbon promises to accelerate climate change, even as humans struggle to curb our fossil fuel emissions.

Few understand this threat better than Zimov. From a ramshackle research station in the gold-mining outpost of Cherskiy, about three hours by speedboat from Duvanny Yar, he has spent decades unearthing the mysteries of a warming Arctic. Along the way, he has helped upend conventional wisdom—especially the notion that the far north, back in the Pleistocene ice ages, had been an unbroken desert of ice and thin soils dotted with sage.

Instead, the abundant fossils of mammoths and other large grazers at Duvanny Yar and other sites told Zimov that Siberia, Alaska, and western Canada had been fertile grasslands, rich with herbs and willows. As these plants and animals died, the cold slowed their decomposition. Over time, windblown silt buried them deep, locking them in permafrost. The upshot is that Arctic permafrost is much richer in carbon than scientists once thought.

Now new discoveries suggest that the carbon will escape faster as the planet warms. From the unexpected speed of Arctic warming and the troubling ways that meltwater moves through polar landscapes, researchers now suspect that for every one degree Celsius rise in Earth’s average temperature, permafrost may release the equivalent of four to six years’ worth of coal, oil, and natural gas emissions—double to triple what scientists thought a few years ago. Within a few decades, if we don’t curb fossil fuel use, permafrost could be as big a source of greenhouse gases as China, the world’s largest emitter, is today.

We aren’t accounting for that. The UN’s Intergovernmental Panel on Climate Change has only recently started incorporating permafrost into its projections. It still underestimates just how wide Pandora’s freezer could swing open—and how much havoc that could unleash.

Permafrost’s potential to warm the planet is dwarfed by our own. But if we hope to limit warming to two degrees Celsius, as 195 nations agreed to during the 2015 Paris talks, new research suggests we may have to cut emissions eight years sooner than IPCC models project, just to account for the thawing that will be going on.

It is perhaps our least appreciated reason to hasten a transition to cleaner energy: To reach whatever goal we set to combat warming, we’ll need to move even faster than we think.

Zimov first came to Cherskiy in the 1970s as a college student to help with mapping on an expedition. He loved the stark landscape and isolation and remoteness from Soviet power centers. The dark winters promised time to think. He returned a few years later and founded the Northeast Science Station, at first under the auspices of the Russian Academy of Sciences. Today he owns and runs it with his son, Nikita. It’s an improvisational operation run on a shoestring and on secondhand equipment. But the station attracts Arctic scientists from around the world.

One day in the summer of 2018, photographer Katie Orlinsky and I joined Zimov in an aging boat to ferry supplies to a carbon-monitoring facility at Ambarchik Bay, near the mouth of the Kolyma on the Arctic Ocean. The site had originally been occupied by a transit station for prisoners bound for Stalin’s gulags, and Soviet-era relics were everywhere. We traversed spongy grasses across a walkway fashioned from a string of old steam radiators. Zimov, bull chested, his long white hair tucked in a beret, probed the ground with a metal shaft as he walked. He’s been doing that a lot lately, to check the depth of the hard permafrost.

Permafrost—ground that remains frozen year-round—is capped by a few feet of dirt and plant detritus. Called the active layer, this soil normally thaws each summer and refreezes in winter, protecting permafrost from rising heat above. But in the spring of 2018, a crew working for Nikita found that dirt near the surface around Cherskiy had not iced up at all during the long dark polar night. That was unheard of: January in Siberia is so brutally cold that human breath can freeze with a tinkling sound that the indigenous Yakuts call “the whisper of stars.” The Soviets used to land heavy planes on the Kolyma. Soil 30 inches down should have been frozen. Instead it was mush.

“Three years ago, the temperature in the ground above our permafrost was minus 3 degrees Celsius [27 degrees Fahrenheit],” Sergey Zimov said. “Then it was minus 2. Then it was minus one. This year, the temperature was plus 2 degrees.”

On one level that’s not surprising. Earth’s five warmest years since the late 19th century have come since 2014, and the Arctic is warming more than twice as fast as the rest of the planet, as it loses the sea ice that helps chill it. In 2017 tundra in Greenland faced its worst known wildfire. Days before we landed in Siberia, thermometers in Lakselv, Norway, 240 miles above the Arctic Circle, recorded a blistering 32 degrees Celsius, or 90 degrees Fahrenheit. Arctic reindeer hid in road tunnels for relief.

Permafrost temperatures globally have been rising for half a century. On Alaska’s North Slope, they spiked 11 degrees Fahrenheit in 30 years. Localized thawing of permafrost, especially in villages where development disturbs the surface and allows heat to penetrate, has eroded shorelines, undermined roads and schools, cracked pipelines, and collapsed ice cellars where Arctic hunters store walrus meat and bowhead whale blubber. Warm summers are already warping life for Arctic residents.

What the Zimovs were documenting in 2018, though, was something different, with implications beyond the Arctic: a wintertime thaw. The culprit, paradoxically, was heavy snow. Siberia is dry, but for several winters before 2018, thick snow had smothered the region. The snow acted like a blanket, trapping summer heat in the soil. At a research site 11 miles from Cherskiy, Mathias Goeckede of Germany’s Max Planck Institute for Biogeochemistry found that snow depth had doubled in five years. By April 2018 temperatures in the active layer had risen 10 degrees Fahrenheit.

The phenomenon wasn’t limited to Siberia. Vladimir Romanovsky, a permafrost expert at the University of Alaska Fairbanks, had for years watched the active layer freeze completely by mid-January at some 180 research sites in Alaska. But as those places also faced a recent period with heavy snow, the freezing slipped first to February, then to March. In 2018, eight of Romanovsky’s sites near Fairbanks and a dozen on the Seward Peninsula, in western Alaska, never fully froze at all.

Globally, permafrost holds up to 1,600 gigatons of carbon, nearly twice what’s in the atmosphere. No one expects all or even most of that to thaw. Until recently, researchers presumed permafrost would lose at most 10 percent of its carbon. Even that, it was thought, could take as much as 80 years.

But when the active layer stops freezing in winter, things speed up. The added warmth lets microbes chomp organic material in the soil—and emit carbon dioxide or methane—year-round, instead of for just a few short months each summer. And the winter warmth spreads down into the permafrost itself, thawing it faster.

“A lot of our assumptions are breaking down,” said Róisín Commane, an atmospheric chemist at Columbia University who tracks carbon emissions by airplane. She and her colleagues have discovered that the amount of CO2 coming off Alaska’s North Slope in early winter has increased by 73 percent since 1975. “We’ve been trying to understand what’s going on in the Arctic by relying on summer,” Commane said. “But after the sun goes down—that’s when the real story begins.”

A few snowy winters don’t make a trend; this past winter there was less snow in Cherskiy, and the soil cooled again considerably. Fairbanks also got little snow. Yet at some of Romanovsky’s sites in Alaska, the active layer again retained enough heat to keep from completely freezing.

“This is truly amazing,” said Max Holmes, deputy director of Massachusetts’s Woods Hole Research Center, who has studied the carbon cycle in both Alaska and Cherskiy. “I’ve largely imagined permafrost thaw as a slow and steady process, and maybe this is an odd five-year period. But what if it’s not? What if things change much more quickly?”

And what if the change becomes self-reinforcing—as it already is, for example, in the case of Arctic sea ice? Sea ice reflects the sun’s rays, keeping the ocean below it cold. But as sea ice melts, the dark ocean absorbs that heat, which then melts more ice.

As a rule, the tipping points at which such feedback loops kick in are tricky to predict. “We know there are thresholds we don’t want to cross,” said Chris Field, director of Stanford University’s Woods Institute for the Environment. “But we don’t know precisely where they are.”

With permafrost, there’s just too much we can’t see. It covers an area more than twice the size of the United States, inhabited by half as many people as New York City, in some of the world’s least accessible terrain. Little of it is monitored directly. Scientists instead study small plots, track others remotely, and draw inferences about the rest—unlike Arctic sea ice, which can be measured in its entirety by satellite. “You can go online and track exactly what happened to sea ice,” said permafrost expert Ted Schuur of Northern Arizona University. “With permafrost, we’re barely looking. We barely have the tools to measure what’s happening.”

One type of permafrost has researchers particularly concerned: the 20 percent or so that contains immense deposits of solid ice. Some of that ice formed when water percolated down through soils and froze as it hit permafrost; some was created over thousands of years during Arctic winters, when the ground contracted and cracked into polygonal patterns. In spring, meltwater filled those crevices, which later refroze. Over time the buried ice grew into massive wedges enveloped by permafrost soil. Duvanny Yar is shot through with them.

Such a structure can unravel swiftly. When permafrost disintegrates, buried ice melts too. As water drains, it transports heat that spreads the thawing, and it leaves behind tunnels and air pockets. The ground sinks to fill those cavities, creating surface depressions that fill with rain and meltwater. The water deepens the pools and chews through their icy banks, until puddles grow to ponds and ponds become lakes. That causes more ground to warm and more ice to melt.

“Abrupt thaw,” as scientists call this process, changes the whole landscape. It triggers landslides; on Banks Island in Canada, scientists documented a 60-fold increase in massive ground slumps from 1984 to 2013. It topples forests. Merritt Turetsky, an ecologist with Canada’s University of Guelph, has tracked abrupt thaw in a black spruce forest near Fairbanks for the past 15 years. Flooding there, she has found, is destabilizing tree roots and trunks. Turetsky suspects all the trees in her “drunken forest” will tip over soon and get swallowed by new wetlands. “There are still little pockets of land, but you have to wade through some pretty wet spots to reach them,” she said.

All permafrost thaw leads to greenhouse gas emissions. But standing water accelerates the threat. The gas that bubbles from the oxygen-deprived mud under ponds and lakes is not only carbon dioxide but also methane, which is 25 times as potent a greenhouse gas as CO2. Ecologist Katey Walter Anthony of the University of Alaska Fairbanks has been measuring the methane coming from Arctic lakes for two decades. Her latest calculations, published in 2018, suggest that new lakes created by abrupt thaw could nearly triple the greenhouse gas emissions expected from permafrost.

It’s not clear how much of this message has reached policymakers. Last October the IPCC unveiled a new report on the more ambitious of two temperature goals adopted at the 2015 Paris conference. The planet already has warmed by about one degree Celsius (1.8 degrees Fahrenheit) since the 19th century. Capping global warming at 1.5 degrees Celsius rather than two degrees, the report said, would expose 420 million fewer people to frequent extreme heat waves, and it would halve the number of plants and animals facing habitat loss. It also might save some coral reefs—and as much as 770,000 square miles of permafrost. But to achieve the 1.5-degree goal, according to the IPCC, the world would have to cut greenhouse gas emissions 45 percent by 2030, eliminate them completely by 2050, and develop technologies to suck huge quantities back out of the atmosphere.

The challenge may be even starker. The 1.5-degree report was the first time the IPCC had taken permafrost emissions into account—but it didn’t include emissions from abrupt thaw. Climate models aren’t yet sophisticated enough to capture that kind of rapid landscape change. But at National Geographic’s request, Katey Walter Anthony and Charles Koven, a modeler at the Lawrence Berkeley National Laboratory, made rough calculations that do add in emissions from abrupt thaw. To halt temperature rise at 1.5 degrees, they estimate, we’d have to zero out our own fossil fuel emissions at least 20 percent sooner—no later than 2044, six years ahead of the IPCC timetable. That would give us just a quarter century to completely transform the global energy system.

“We’re facing this unknown future with an incomplete set of tools,” Koven said. “The uncertainty isn’t all on our side. There are a lot of ways things could turn out worse.” There’s more than one way to make new lakes, for example.

A few weeks after leaving Siberia, Orlinsky and I took a raft trip through Alaska’s Gates of the Arctic National Park with ecologist Ken Tape, a colleague of Walter Anthony’s at the University of Alaska. A floatplane dropped us and river guide Michael Wald at Gaedeke Lake, in the central Brooks Range. From there we made our way south down the Alatna River. September sun danced on the water. Within a mile or so we found chewed sticks along the bank. We’d been on the river a week when we arrived at a 38-acre lake that hadn’t been there before. At its center was an enormous beaver lodge.

Tape has been using aerial and satellite photographs for years to track how plants and wildlife are changing in Alaska—and how that might affect permafrost. As permafrost thaws and the growing seasons lengthen, the Arctic is greening: Shrubs in Alaska river plains, for example, have nearly doubled in size. (While vegetation growth will take up more carbon, a 2016 survey of experts concluded that Arctic greening won’t be nearly enough to offset permafrost thaw.) The vegetation is drawing animals north.

With willows now tall enough to poke through snow, snowshoe hares can find winter food and hiding spots all the way to the Arctic Ocean. Typically forest dwellers, they’ve now colonized Alaska’s North Slope, hundreds of miles from any real forest. Lynx, which prey on hares, appear to have followed. Both are probably traveling a trail blazed by moose, which also eat willows and now number roughly 1,600 along the Colville River, where they were absent before.

Those discoveries led Tape to search photographs for other tundra newcomers. “As soon as I thought about beavers, I seized on it,” he said. “Very few species leave a mark so visible that you can see it from space.”

In images from 1999 to 2014, covering just three watersheds, he spotted 56 new beaver pond complexes that hadn’t been there in the 1980s. The animals are colonizing northern Alaska in earnest, moving at about five miles per year. Tape believes there are now up to 800 beaver pond complexes in Arctic Alaska, including the one with the massive lodge on the Alatna. Tape dubbed it Lodge Mahal.

It was quite a sight: a mound of branches and saplings, about eight feet high by 35 feet across, plastered with mud and moss and sitting in a waist-high lake surrounded by marsh. The water had been diverted from the river by a series of dams. “That entire swamp around Lodge Mahal is new,” Tape said. “If you went back 50 years, there’d be zero beavers here.”

Tape and Wald had wanted to explore the Alatna in part because a guide who works for Wald had earlier found beaver-chewed wood along the Nigu River. The Nigu starts near Gaedeke Lake, the Alatna headwaters, but on the other side of the Continental Divide—and so it flows north into the Colville River and the Arctic Ocean. Along the Alatna, above Lodge Mahal, we found other ponds and abandoned dams. Tape now thinks that beavers are on their way to the North Slope, and that they’re using the Alatna as a route through the Brooks Range. “We’re seeing this expansion in real time,” he said.

He can’t prove that climate change is driving it; the beaver population also has been rebounding since the end of the fur trade, a century and a half ago. But in any case, the bucktoothed engineers could significantly remake permafrost landscapes. “Imagine if you were a developer and you said, I’d like permission to put three dams on every other stream in the Arctic tundra,” Tape said. “That’s what this could be like.”

Tape has seen a preview. Southeast of Shishmaref, on Alaska’s Seward Peninsula, photos of a tributary of the Serpentine River show no change at all between 1950 and 1985. By 2002 beavers had moved in and flooded the landscape. By 2012 some ground had collapsed and become wetlands. Permafrost was on its way out.

A few hundred beavers won’t reengineer the Arctic. But the animals may be heading north in Canada and Siberia too, and they reproduce quickly. Argentina’s experience is instructive: Twenty beavers were deliberately introduced in the south in 1946 in order to foster a fur trade. Today that population hovers around 100,000.

In the Zimovs’ vision of the past and future of Arctic permafrost, wild animals also play a central role—but the beasts are bigger than beavers, and their effect on permafrost more benevolent. The herds of bison, mammoths, horses, and reindeer that lumbered across the Pleistocene steppes, Sergey Zimov has long argued, did more than just eat the grass. They maintained it. They fertilized it with their waste and packed it down, trampling mosses and shrubs and ripping out tree saplings.

Since the last ice age, those dry, rich grasslands have been replaced in eastern Siberia by damp tundra, dominated by mosses in the north and forests farther south. One key driver of that change, according to Zimov, was human hunters who decimated the herds of large grazers, by about 10,000 years ago. Without grazers to fertilize the soil, grasses withered; without grasses to soak up water, the soil got wetter. Mosses and trees took over. But if humans hadn’t pushed the ecosystem beyond a tipping point thousands of years ago, there would still be mammoths grazing in Siberia.

Almost 25 years ago, on lowlands near Cherskiy, Zimov created a 56-square-mile demonstration project called Pleistocene Park. His idea was to bring large grazers back and see whether they would bring back the grasslands. He and, eventually, Nikita fenced in wild horses and later trucked in yaks and sheep from Lake Baikal. This past spring Nikita hauled in 12 bison from Denmark, traveling 9,000 miles across Russia by truck and barge. In 2018 the Zimovs joined forces with Harvard University geneticist George Church, who thinks he essentially can clone a mammoth. The hope is that one day those now extinct beasts will be stomping around Pleistocene Park, thriving in the Anthropocene.

The park is the ultimate test of Sergey Zimov’s hypothesis—and, he hopes, a hedge against future climate change. Grasslands, especially when snow covered, reflect more sunlight than does dark forest. Grazing animals tamp down deep snow, allowing heat to escape the soil. Both things cool the land. If wildlife could restore grasslands, it would slow permafrost thaw and thus climate change. To make a real difference, though, you’d need to unleash thousands of zoos’ worth of animals across millions of acres of the Arctic.

The Zimovs say the evidence from their 36,000-acre park is promising. Even with only about a hundred animals, the park’s grasslands stay substantially cooler than the ground in the surrounding area.

The gap between the Zimovs’ ambitions and the reality of the park is unquestionably large. During a tour one afternoon, Orlinsky and I hiked soggy grasses to a stretch of marsh to watch the horses. A lone bison hid in the distance. Nikita loaded us onto an eight-wheel mini-tank and took us crashing through the willows. After a steep climb we plowed over some skinny larches. This is why he needs giant herbivores, Nikita said: “At the moment I don’t have any animals which can kill those trees.” He spends a lot of time raising funds, most recently in California, hobnobbing with the likes of former Governor Jerry Brown, just to keep this proof of concept going.

The concept has its critics. Some scientists dispute the Zimovs’ estimates of how many large animals were roaming around Siberia in the Pleistocene, or insist that their theory of ecological change, both past and present, is too simplistic. Above all, most criticism seems leveled at the Zimovs’ audacity. Max Holmes of Woods Hole, who knows them well, sees a spark of genius in their work. The Zimovs are “at the fringe,” Holmes said, “but that’s often where big ideas and big changes originate.”

Outside Pleistocene Park, the modern world has responded to the warming Arctic with complacency. We’ve spent decades ignoring the evidence of climate change and hoping that things won’t get too bad. We count on technological advances that seem always just out of reach. And we do this in spite of the fact that climate scientists—permafrost experts in particular—say all signs point to the need for urgent and even audacious action.

The Zimovs are different: They’ve spent their lives battling an unforgiving landscape that rewards bullheadedness. Is trying to save permafrost by restoring the Arctic steppe, they ask, really so much crazier than counting on humans to quickly retool the world’s energy system? Maybe we need a little craziness.

“Fighting climate change needs multiple actions from multiple different fronts,” Nikita said. Only if we combine them all can we make the future “not entirely miserable.”