A thousand feet above the glistening, iceberg-dotted water of the ocean off of East Greenland, oceanographer Josh Willis braces for balance, his feet spread wide on the metal floor of a specially-outfitted airplane. He grips a wide grey cylinder, hovering it over a 6-inch-wide bottomless tube.

The pilot’s voice crackles over the intercom: “3, 2, 1, zero, DROP.”

Willis lets the cylinder go. With a whoosh, it slips down the tube and into the wide-open air.

The plane banks hard to the right and everyone on board rushes to a window. “I see it!” yells Ian Fenty, another oceanographer on the project, as the probe—designed to sink to the seafloor and record the properties there—splashes down.

Willis, Fenty, and a crew of other scientists and pilots are flying the edge of Greenland’s vast ice sheet to figure out how the ocean eats away at the ice, speeding or slowing its slide into the water, where it melts, raising sea levels worldwide.

But exactly how much ice it will deposit, and how fast, is still an open question. Greenland is currently the biggest contributor to global sea level rise. By 2100, will its ice sheet’s melt add inches to the world’s oceans—or will it add much more?

That’s a trillion-dollar question. Nearly 70 percent of Earth’s population lives within 100 miles of a coast, and vast amounts of infrastructure—from airports to ports to cities to roads to Internet cables—sits in zones that could flood within decades. Small, low-lying island nations, city planners, insurance adjustors, homeowners—everyone is clamoring for the most accurate estimates of how much extra water they’ll need to prepare for.

And for that, says Willis, they need to know what happens here, where ocean meets ice.

“This is where it all happens,” he says. The flooding of the future is being defined here and now, in the glittering sea below.

A sudden lurch into melting

Greenland’s ice is shrinking, this we’ve known for a while, since the science of global warming, a famous climate scientist likes to say, is older than the technology that makes our iPhones fast and the Internet run smoothly.

But until the 1990s, the ice in Greenland was remarkably stable, even as air temperatures rose because of human-caused climate change. Each year, the ice sheet lost some weight as ice flowed like taffy from the center of the ice sheet, through funnel-like outlet glaciers at its edge, spilling into the ocean. But enough snow fell on top of the mile-high interior of the ice sheet to balance out the losses.

In the 1990s, scientists thought that the big ice sheets in Greenland and Antarctica responded slowly to changes in climate, shuddering into motion like bears waking up from hibernation. Yes, they’d respond to the human-caused climate change that was gripping the planet, the thinking went, but it would take decades or even centuries to really see the impacts.

“Early on, we weren’t thinking about Greenland as being really critical on these kind of decadal scales, and we didn’t have tools to look at them on those time scales,” explains Twila Moon, a glacier expert at the National Snow and Ice Data Center.

But around 1997, something changed. Scientists studying Jakobshavn glacier, on Greenland’s western coast, watched in alarm as a tongue of ice that had for years poked out into a fjord started to shrink. The tongue was about 15 kilometers long in 1997. By the early 2000s—a scant half decade later—that tongue was gone.

“We suspected that this could happen from time to time, but this was the first time we’d seen anything like it,” says David Holland, who led the team studying the rapid disintegration of the ice tongue.

Today, the Greenland ice sheet is losing mass about six times faster than it was just a few decades ago, whatever tenuous balance that existed before long since upended. Between 2005 and 2016, melt from the ice sheet was the single largest contributor to sea level rise worldwide, though Antarctica may overtake it soon.

Within the past 50 years, the ice sheet has already shed enough to add about half an inch of water to the world’s oceans, and that number is increasing precipitously as the planet heats. During this summer’s extreme heat wave that parked over Greenland for a week and turned over half its surface ice to slush, meltwater equivalent to over 4 million swimming pools sloughed into the ocean in a single day. Over the month of July, enough melt poured into the ocean to bump sea levels up by an easily measurable half a millimeter.

Overall, there’s enough water locked up in the Greenland ice sheet to add about 25 feet to the world’s oceans. It’s not likely that such catastrophic loss will happen soon, as in within the next few hundred years. But the whole of the ice sheet doesn’t have to collapse to cause massive, planet-wide reverberations.

“When I started this research, I never would have guessed that warm subsurface waters could unravel an ice sheet,” says David Holland, an oceanographer at NYU. “But it’s becoming clear that they can, and that they are.”

The ice tongue’s retreat started scientists on a Sherlock Holmes-esque quest for the culprit.

At the time, most models of ice sheet dissolution more or less assumed that the ice melted from the top, when warm pockets of air parked themselves over the sheet. But it hadn’t been exceptionally warm when Jakobshavn suddenly retreated. That meant there must be another factor at play.

One team of oceanographers, led by David and Denise Holland, had a collective hunch. Maybe, they thought, something changed in the water upon which the ice tongue had been floating. If the ice didn’t melt because of something warm above it, maybe it melted from below, like an ice cube in a glass of water.

The problem was that Greenland’s winding, remote, ice-choked coastline is over 27,000 miles long—and there were only a few spots measured consistently. By comparison, California’s coast, about an eighth as long, has hundreds of buoys beaming out information constantly.

But luckily, the Danish fisheries service had been steaming through the fjord system around Jakobshavn for years, testing water temperatures and other data that helped them understand what kind of conditions were good for fish.

The Holland’s team pieced together bits of that data and discovered that the water in the fjords started to get warmer just when the Jakobshavn ice tongue started to retreat.

So the culprit wasn’t just hot sun beating down on the surface of the ice, though that was certainly part of the problem. It was also a long, steady stream of warm ocean water that suddenly managed to reach Jakobshavn.

But where did that water come from? And what would it do into the future? The questions piled up, and teams of scientists flocked to Greenland to figure it out.

Back to the mystery

After the Jakobshavn discovery, it became clear that slight changes in the ocean waters here—their temperature, saltiness, or currents—could affect the ice anywhere they made contact.

In the long term, that’s bad news. The ocean has done the lion’s share of soaking up the excess heat already trapped in the atmosphere by human-caused climate change, absorbing over 90 percent of all the extra warmth since the Industrial Revolution began. Over time, there will likely be more and more warm water available to melt ice.

And there is plenty of ice to be melted. The snouts of some 200 Greenland glaciers poke out over the water, making them vulnerable to the same kind of oceanic edge-nibbling. Recent modeling done exclusively for National Geographic shows that if climate change continues unabated, the big glaciers—like Jakobshavn, or Helheim Glacier on the southeastern Greenland coast—would shed enough ice to add about one centimeter to the seas. Recent research also suggests that by 2100, the enough of the island’s ice could slip into the sea (either as bergs or as meltwater) to send sea levels up 2 to 12 inches. By 1,000 years out, if greenhouse gas emissions continue unabated, the ice could be gone.

The coastal glaciers are like rootlets tendriling off the taproot of Greenland’s interior, the channels through which much of the ice is lost. Ice from the interior flows out through these glaciers, eventually reaching the ocean, where it melts and contributes to sea level rise.

Over the first decades of this century, scientists developed ever more sophisticated ways to measure how fast the ice was flowing. They looked at the bright white outlet glaciers with radar and visible-spectrum satellite images; they used a powerful pair of satelliites to “weigh” the ice loss as it was happening; and they built models that captured the great groaning and strain of the ice. When they got a better view of how much of ice was getting lost, they could see that many of those 200 outlet glaciers were shrinking—just like Jakobshavn.

Maybe, Willis and his colleagues hypothesized, the answer was, once again, the ocean.

So they schemed up a project. They’d need to know a few things: what was happening with the ice on land (fine, they’d fly a plane over the ice margin and track its activity with radar). They needed to know what the bathymetry underneath the ice looked like, both on land and in the water. Was the ice reaching deep down, like a fat wine cork stopping up the fjord? Or was it a thin layer? (OK, they’d send ships up into the fjords to measure the bottom depth).

And finally, they needed to know what was going on with the water itself. Was it warm? Where? And why? For that, they’d have to get some sensors in the water—by dropping them, they decided, from a fixed-up, tricked-out DC-3 plane originally built in 1942.

In 2015, they powered up their planes and boats and got to science-ing. A picture quickly came together. “When we started looking, we saw these fjords that were 200, 400, 800 meters deep!” exclaims Fenty.

And the water around the coast? Parts of it were often shockingly warm—sometimes up to 10 degrees Celsius or more, well above the just-above-freezing temperatures they expected, he says.

That setup creates a perfect environment for melting ice. In this part of the ocean, water at the surface tends to be chilly and fresh. But if you dived down a few hundred feet, you’d hit a warm, salty layer—a current, part of the Gulf Stream, that comes straight from the tropics with the warmth of the equatorial sun in its watery bonds.

Much of the time, the Greenland coast is insulated from that warm water. A wide shelf around the edge of the island acts like a seawall blocking the water from coming in contact with the ice. But sometimes, when long-term weather patterns slip into a particular mode, it can spill over the wall. And once it’s there, it can spill into the deep fjords. And once it’s in the fjords, it can get all the way to the ice front.

At the ice front, the glaciers fill the deep fjords like huge ice corks in a bottle. But the corks are delicate. Each hit of warm water eats away at them a little more. Just a few days later on that research junket, the scientists watched a particularly dramatic example. As they flew low over the leading edge of the massive Helheim glacier, aiming to drop a probe through a hole in the in the mélange of giant bergs floating at the glacier’s snout, they saw water roiling up through the hole “like a bubbling cauldron,” says Willis.

When the probe pinged back data, it showed a warm wall of water extending straight down 2,000 meters to the bottom of the fjord: A solid wall of water ready to melt the glacier.

Each year since 2015, the team has dropped about 250 probes into the ocean around the edge of Greenland. They’ve found the toasty water nosed up to the end of glaciers around the island most of the time in most of the places.

And the exceptions, they think, prove the rule. At Jakobshavn, for example, they saw the water cool down for a few years—and the glacier responded in kind, slowing its retreat. But that just strengthened their hypothesis, says Willis, making it even clearer that the ocean acts as the “main control on that glacier.” The system is at the whim of water.

The future of the ocean

On the plane, the data starts beaming back to Fenty’s screen a few minutes after the drop. Near the surface, the ocean is just barely above freezing. But as the probe sinks—100 meters, 200, 400, 500—the familiar signal appears, indicating warm water.

“There it is,” says Fenty. “It’s there again, that Atlantic water.”

The warm layer has intruded upon most of this part of the coast for most of the years the team has been dropping probes. But that’s partly a product of a weather cycle that influences the wind and ocean currents around the island. Currently, the cycle is in a phase that lets warm Atlantic water slosh toward Greenland instead of being pushed cleanly toward Europe. When the phase flipped briefly, the currents around the island cooled down—and hence Jakobshavn slowed its melting.

Changes to the cycle, like many other atmospheric and oceanic patterns, hasn’t been directly linked to climate change yet. But there are some hints that the phase that causes warm water to get up to the edge of the ice is getting more prevalent.

“I think the bottom line is, in a warming planet, we expect the glaciers to retreat and the ice sheet to change, a lot,” says Fiamma Straneo, an oceanographer at Scripps Institute of Oceanography. “Greenland tends to be an integrator of the climate signal. What we’re seeing is the effects of a warmer atmosphere over the Arctic—as well as probably a warmer ocean.”