The last woolly mammoth on Earth did not die at the end of the Ice Age. It died on a scrap of tundra in the Arctic Ocean, roughly nine thousand years after the ice retreated from Europe and something like six centuries after Egyptian crews finished the Great Pyramid at Giza. By then the mammoth was already a fossil animal across the whole northern world. Everywhere a person could have walked, they were gone. Everywhere except one island that a rising sea had turned into a trap.
The island that closed behind them
For most of the last glacial period, Wrangel Island was not an island. It was high ground on a vast plain, part of the mammoth steppe that ran from Europe across Siberia into what is now Alaska, all of it laced together by land that sat above a lower sea. Mammoths grazed across it in the tens of thousands.
Then the ice melted and the water came up. Around ten thousand years ago the sea flooded the low country and closed the gap between Wrangel and the Siberian mainland, leaving a block of tundra roughly seven and a half thousand square kilometers, adrift in the Chukchi Sea. Whatever mammoths happened to be standing on the high ground when the water rose were now cut off. Genomic work later put the founding herd at no more than eight breeding animals — a rounding error of a population, the kind of number that usually precedes an extinction rather than a six-thousand-year reprieve.
On the mainland, the story was already over. The great herds had thinned and vanished by about ten thousand years ago, under some combination of a warming climate and human hunters. A relict population held on into the mid-Holocene on St. Paul Island, in the Bering Sea, and died out there around 5,600 years ago when its freshwater ran short. After that, Wrangel was the only place on the planet where a woolly mammoth still drew breath.
Six thousand years on a stranded rock
Eight animals should not have been enough. Small isolated populations tend to spiral: inbreeding stacks up harmful mutations, fertility drops, and the line collapses within a few generations. The Wrangel herd did the opposite. Within about twenty generations it had climbed to somewhere between two hundred and three hundred mammoths, and there it stayed. For roughly six thousand years — more than two hundred generations — the island carried a small, steady herd of full-sized woolly mammoths while the rest of the world moved into the age of agriculture, cities, and writing.
The arithmetic of that is worth sitting with. When the pyramids at Giza rose, mammoths were alive. When Egypt was a going concern with a bureaucracy and a calendar, mammoths were alive. They were still walking the tundra of Wrangel when the Great Pyramid was already an old monument, its casing stones weathering in the sun. The animal that stands in every museum diorama beside a spear-throwing hunter in furs was, in fact, a contemporary of recorded history.
Isotopes locked in their bones sketch a surprisingly ordinary life. Carbon and nitrogen readings from the island’s mammoths stayed stable straight through to the end, without the dietary stress that showed up in mainland herds before they died. Analysis of the bones even suggests the Wrangel animals leaned less on their fat reserves than their Ice Age Siberian ancestors had, a sign of a milder island climate. By every measure researchers can pull from the ground, these were not desperate animals clinging on. They were doing fine.
The vanishing
Then, around four thousand years ago, they were simply gone.
There is no long tail in the record, no slow guttering out over centuries. The most recent radiocarbon-dated mammoth bones on Wrangel cluster near 2000 BC, and then nothing. A demographically stable population that had held its numbers for six millennia disappeared in what, on a geological clock, reads as an instant.
For a long time the assumed cause was genetic — a “genomic meltdown,” the idea that all those generations of inbreeding finally cashed out in a population too broken to continue. In 2024 a team at the Centre for Palaeogenetics in Stockholm, led by Love Dalén with Marianne Dehasque as first author, tested that idea directly. They sequenced twenty-one mammoth genomes — fourteen from Wrangel, seven from the mainland ancestors — spanning fifty thousand years of the species’ existence, and watched the island population’s DNA change in near-real time.
The genomes did show the scars of the bottleneck. Diversity was low, inbreeding was real, and the mammoths had lost variation in the major histocompatibility complex, a cluster of immune genes that helps a body recognize disease. But the herd was also quietly cleaning house. Over the millennia, natural selection was purging the most dangerous mutations from the population even as milder ones accumulated. The diversity that remained was declining, but slowly — nowhere near a crash. Dalén’s conclusion was blunt: the population was not simply too small, and it was not doomed to die out for genetic reasons. As Patrícia Pečnerová, a co-author on the mammoth genome work, put it, the story scientists had been telling themselves was wrong.
Which leaves the mystery larger than before. If genetics didn’t kill them, and the climate didn’t obviously turn against them, and the isotopes say they were eating well up to the end, then something else came through and finished a healthy herd in a hurry.
What came through the door
The candidates are all short, sharp events, because a slow one would have left a trace. An earlier study led by Laura Arppe of the Finnish Museum of Natural History found sulfur climbing through the island bones right to the end, a hint that weathering bedrock in the central mountains was fouling the island’s drinking water. On top of that, a population weakened at the margins is exactly the kind that a single freak season can erase. Hervé Bocherens, one of that study’s authors, put it as a stacking of pressures: a herd already worn down by bad water and thin genetics could have succumbed to something like an extreme weather event. The genome work that came after narrowed what that first pressure could mean — the herd was inbred, but not failing — and left the last step standing on its own. A rain-on-snow event, the icing storm that glazes tundra in a shell too thick to dig through, could cut off the food supply long enough to take a herd of a few hundred down to zero.
Humans are the usual suspect for a megafauna extinction, and here they have something close to an alibi. The earliest sign of people on Wrangel is a short-term hunting camp at Chertov Ovrag, on the southern coast, and its dates fall between roughly 3,650 and 3,350 years ago — a window that opens at least two centuries after the youngest dated mammoth. What those people ate was seal, walrus, and above all birds, snow geese by the dozen; nothing in the camp says anyone there hunted a large land animal. The gap is a real one, but it is an absence of evidence rather than a proof: Arppe’s team noted that earlier visitors who left no trace cannot be ruled out, only that nothing yet found puts a human and a mammoth on the island at the same time.
So the most complete population history of any extinct animal ends on a shrug. The genomes are sequenced, the isotopes are measured, the bones are dated, and the last woolly mammoths still vanish through a door science can describe but not quite name. What the record shows clearly is only the shape of the ending: not a long decline, not a doomed inbred remnant wheezing out, but a small herd of mammoths, healthy by every measure that survives, standing on their island a thousand years into the Bronze Age — and then not there at all.