How scientists unraveled the cause of the devastating flood in Tibet and Nepal

Kristen Cook, a geomorphologist at Grenoble Alpes University, is a night owl. When a phone call from a colleague woke her up at 10:45 a.m. Central European Time on Wednesday, 26 August, she had already missed a dozen emails and streams of text messages alerting her to news of a devastating flood along the Lhende Khola River in Tibet and Nepal. Still in pajamas, “I got up and sat at my computer,” she recalls. “And I did not get up again for 5 hours.”

Cook, along with hundreds of scientists around the globe, has been working tirelessly to piece together the sequence of events that led to the massive flood, which has killed nearly 600 people, left nearly 2000 missing, and destroyed numerous towns and hydropower plants along the Nepal-Tibet border. Researchers tell Science it was a process challenged by some early missteps, incomplete data, and frustrating conditions, including clouds that obscured satellite imagery. But although many questions remain—including about the origins of the floodwaters themselves—the global scientific community cooperated to unravel the mystery “ridiculously fast,” says Göran Ekström, a seismologist at Columbia University’s Lamont-Doherty Earth Observatory.

The first accounts, including from the U.S. Geological Survey (USGS), indicated the event was a magnitude 4.4 earthquake, which might have shaken rock and ice loose in a landslide that sparked the flood. A dam break was also touted as a possibility. But within 15 minutes of waking up, Cook, who has been conducting research on the Lhende Khola, had downloaded publicly available seismic data from a Nepali station located north of Kathmandu to see for herself. 

“I could already see that some of the initial attribution was wrong,” she says. 

Rather than the quick, intense seismic jolt typical of an earthquake, the data showed low-frequency, long-period waves characteristic of prolonged rumblings: a giant landslide, which itself generated the seismic signal.

On the opposite side of the Atlantic Ocean from Cook, Ekström was using data from the Global Seismographic Network, a collection of about 200 seismometers operated by many countries, to come to the same conclusion. He fed the seismic data into algorithms developed at Lamont-Doherty to find that a more accurate magnitude was 5.7—equal to a significant earthquake. It soon became “clear that [the Lhende Khola landslide] was one of the largest in the last decade or two in the world,” he says.

Cook, Ekström, and a smattering of other top natural hazard researchers soon created a Slack group to share data, discuss interpretations, and argue. (As of this publication, the group has more than 85 scientists.) It was here that the next useful observations got shared: satellite imagery from Planet Labs, which takes daily images of Earth’s surface at a resolution of 3 to 5 meters. But there was a problem: In the few available images of the mountains where the seismic signal originated, there were “very strategically placed inconvenient clouds,” Cook says. What could be clearly seen was a sharp edge, at about 5200 meters above sea level, where a glacier that blanketed the mountain had snapped off.

The imagery soon led to reports that the explanation was a glacial collapse: A thick layer of ice had slid off the underlying rock, potentially because of warming at its bottommost layer, until it crashed into the valley below. There, the shattered ice and rock flowed down the Lhende Khola valley at speeds approaching 150 kilometers per hour, kicking off a flood that caused water levels to rise as much as 9 meters in 30 minutes.

But Ekström wasn’t convinced. He ran calculations of how much falling mass was needed to create the magnitude 5-plus seismic signal. The answer was hundreds of millions of tons—far more than contained in the glacier alone. At the same time, Cook says, cryosphere scientists calculated the energy that should have been released when the glacier collapsed, which in turn could have caused the ice to liquefy and contribute water to the flood. The energy was only enough to melt a small fraction of the ice, they found—not enough. For a pure glacial collapse, “The physics don’t work,” Cook says.

The mystery could only be fully resolved with cloud-free satellite imagery. And by Wednesday evening, it arrived, both from a satellite operated by the Indian Space Research Organisation and Landsat-9, a joint NASA-USGS satellite. Now, the underlying mountain was visible—and it revealed a huge rock landslide had indeed occurred. “The glacier just came along for the ride,” Ekström says.

Just a few years ago, reaching that kind of conclusion “would have taken months or years … because we didn’t have the imagery to piece this together,” Ekström says. This time, the process took less than a day.

Ironically, piecing together the puzzle was more difficult for scientists near the scene in Nepal, which lacks its own satellites. Although social media videos circulated quickly, “The major obstacle was that we didn’t have any ground information for more than 3 hours,” says Adhikari Basanta Raj, director at the Centre for Disaster Studies at Nepal’s Tribhuvan University. But drone footage acquired the day after the flood eventually confirmed the landslide narrative, Raj says. He calculated that the energy released during the landslide was larger than the atomic bomb used in Hiroshima, Japan.

From there, the story gets a little murkier. The event has been called a “cascading hazard” for the way it strung together multiple events: a landslide, debris flow, and flood. Similar sequences have played out in the Himalayas before, notably in the 2021 Indian Chamoli disaster that killed 200 people and damaged two hydropower plants, as well as Nepalese events in 2012, ’15, and ’17. But the magnitude of the newest event was 10 times larger, Cook says.

One mystery is how so much water got mobilized into the flood. Some researchers initially suggested a glacial lake outburst flood (GLOF), in which the “liquid concrete–like slurry” of the fallen mountainside formed a temporary dam on the Lhende Khola, then rapidly released millions of cubic meters of water when the dam burst. But the seismic and satellite data show “absolutely no evidence” of a GLOF, Cook says. Instead, she suggests the flow of rock, ice, and sediment scooped up river water as it rushed downstream. There may also have been large volumes of old ice on the valley floor that got pulverized, melted, and incorporated into the flow.

To truly understand the cascade of hazards, a mix of helicopter reconnaissance, on-the-ground investigations, and talking to local people will be required, says Alton Byers, a mountain geographer at the University of Colorado Boulder. He is traveling to Nepal in September for research on preparing communities for glacial hazards.

Perhaps most significant for future disasters is the possible connection to global warming. Byers points a finger at the gradual thawing of high-altitude permafrost, which binds rock and ice together. Other researchers hope to assess whether snow melted in the days leading up to the event, causing water to seep into and destabilize the rock. Furthermore, if the glacier was buttressing the mountain, any melt could have removed an important support system. 

“Landslides have been occurring as long as the mountains have been building,” Ekström says. But “in areas where glaciers are disappearing, these landslides are occurring more frequently.”

Now, in Nepal and Tibet, the attention is focused on recovery efforts—and on how to improve early warning systems for downstream areas in the future. 

“I know these villages, and to see what’s happened is unbelievable,” says Cook, who was most recently at the river in May. “To watch these videos … then to be working on something that’s so tragic is hard.”