A reconstruction¹ of the the 2023 Sikkim floods has revealed that decades of glacier retreat and permafrost thaw set the stage for catastrophe — and could trigger another disaster at any time.

On October 3, as heavy rains pounded the northeastern Indian state, a massive chunk of frozen moraine — 14.7 million cubic meters of rock, ice, and debris — collapsed from the flanks of a retreating glacier into South Lhonak Lake. The impact unleashed a towering 20-meter wave, breaching the lake’s natural dam and releasing nearly half its volume — 50 million cubic meters of water — downstream.

The resulting Glacial Lake Outburst Flood (GLOF) proved far more destructive than models had predicted, scouring away roughly 270 million cubic meters of sediment, five times the volume of water released. The flood killed 55 people, displaced thousands, and battered the 1,200-megawatt Teesta-III hydropower dam.

"Current conditions would allow another event,” said Dave Petley, an earth scientist at the University of Hull and co-author of the study. "The lake still holds a large volume of water, and the surrounding moraine remains unstable.”

Petley is part of an international research team led by Ashim Sattar, a glaciologist at the Indian Institute of Technology, Bhubaneswar. Using high-resolution satellite imagery, seismic data, meteorological records, field observations, and numerical modeling, the scientists found that existing risk assessments, including Sattar’s own 2016 model, had underestimated the flood’s power, particularly the impact of debris-laden flows.

“The sheer volume of sediment carried downstream was unexpected,” Sattar said. “It supercharged the flood.”

Predicting GLOFs

As India moves to rebuild the Teesta-III dam, experts warn that stronger hazard assessments and early warning systems (EWS) are crucial.

“A rainfall-based warning system, like Hong Kong’s, could help predict landslides using radar and rain gauges,” said Petley. “Ground deformation monitoring could further enhance early warnings.”

For an EWS to be truly effective, it must go beyond lake-level monitoring, said Rakesh Kumar Ranjan, a hydrologist at Sikkim University. In glacier-rich regions like Sikkim, real-time stream monitoring offers a cost-effective way to detect sudden surges. Any system, he stressed, must also include evacuation plans, response training, and public awareness.

Predicting a GLOF remains complex. “A lake can burst in many ways,” said Sonam Wangchuk, a cryosphere specialist at the International Centre for Integrated Mountain Development (ICIMOD). Rising water volume can weaken a moraine wall, while earthquakes or heavy rain can trigger landslides, displacing water and causing an outburst. When these factors interact — as in Sikkim — prediction becomes even harder.

Floods are changing, and time to adapt is running out, as highlighted in a study² of 1,015 floods in the high mountains of Asia. “A single monsoon cloudburst or glacial collapse can spark cascading disasters, overwhelming unprepared regions,” Wangchuk warned.

South Lhonak: A ticking time bomb

South Lhonak Lake’s rapid expansion over the past two decades made the disaster inevitable. First observed in the 1960s, its growth rate doubled since 2004, from 0.023 km² to 0.046 km² per year. Glacier retreat and thinning, driven by long-term climate warming, fueled this dramatic expansion, said Sattar.

Average annual temperatures in the region have risen by 0.08°C per decade since 1951. Record summer heat in 2020, 2022, and 2023 accelerated glacier mass loss, weakening the frozen moraine that once acted as a natural dam. “We found abundant permafrost around the lake and in the lateral moraine that collapsed,” Sattar noted.

Beyond warming, internal glacial processes hastened the lake’s growth. “Even with year-to-year climate variability, once a glacial lake forms, it can keep expanding independent of direct warming,” he said.

These underlying shifts turned the October 3–4 rainfall into a trigger, unleashing a chain reaction. Heavy rains saturated slopes, increasing landslide risk. Within 18 minutes of the moraine collapse, floodwaters tore through an Indo-Tibetan Border Police camp 7 km downstream. Two hours later, the flood reached the Teesta-III hydroelectric dam in Chungthang, 68 km away.