Unstable dams in the sky
In the Shaluli Mountains, a rugged part of the vast Hengduan range on the Tibetan Plateau, a hazardous geological process is accelerating. As global temperatures rise, glaciers are retreating at rates described as exceptional—in some parts of the Himalayas, the pace of ice loss is at least ten times higher than the average over past centuries. This rapid meltwater production leads to the formation of thousands of new glacial lakes, many of which are held back only by unstable dams of loose rock and soil called moraines.
These moraine dams are inherently unstable. Composed of poorly sorted and unconsolidated sediment left behind by retreating glaciers, they are highly susceptible to failure. Some dams also contain massive cores of buried ice; as this ice melts, it can weaken the dam's internal structure from within, leading to subsidence and eventual collapse. A Glacial Lake Outburst Flood (GLOF) occurs when one of these dams breaches. Triggers can vary, from a large wave caused by an ice or rock avalanche plunging into the lake, to simple overtopping from intense rainfall or snowmelt. The result is a sudden, violent release of enormous volumes of water and debris that can devastate downstream areas. Across the Himalayas, more than two dozen GLOF events have been recorded since 1935.
Forecasting a deluge
Given the threat to communities and infrastructure, scientists are developing sophisticated methods to monitor these high-altitude lakes and predict which ones pose the greatest risk. This uses a combination of remote observation and direct measurement. Researchers use satellite imagery, such as that from the Landsat program, to conduct large-scale surveys, tracking the number and size of glacial lakes over time. Studies in the Chinese Himalayas show that the total area of moraine-dammed lakes expanded by over 25% in two decades. This remote tracking allows scientists to create inventories of potentially dangerous lakes and flag those that are expanding rapidly.
For a more detailed risk assessment of high-priority lakes, scientists conduct bathymetric surveys. This involves mapping the lake floor to calculate its precise volume—an important variable for flood modeling. In these remote, hard-to-reach locations, researchers may use equipment like sonar deployed from uncrewed surface vessels to gather this data. The bathymetric data, combined with information on the moraine dam's structure and the surrounding topography, is fed into computational models. These models simulate GLOF scenarios, predicting the potential flood path, discharge volume, and downstream arrival times. This information is important for developing early warning systems, which can provide precious minutes of notice to downstream populations when a breach is imminent.