“The melting Himalaya” covers four distinct processes, each with its own timescale, its own evidence and its own hazards. Keeping them apart is the first step towards understanding why some mountain disasters are tied to warming and others are not.
When a Himalayan glacier collapses, a lake bursts or a slope fails, the explanation offered is often the same: the ice is melting. The phrase is not wrong, but it hides four separate processes that behave differently, are measured differently and threaten different things. Glaciers, seasonal snow, permafrost and glacial lakes respond to the same warming on different clocks, and a hazard that involves one of them may be untouched by another.
Glaciers: slow accounts, long memory
A glacier is an account balanced over years. Snowfall on its upper reaches adds ice; melt on its lower reaches removes it. When losses exceed gains the glacier thins and its snout retreats, but with a delay that can run to decades, so the ice visible today reflects past climate as well as present. At Langtang Lirung, an international analysis published in September 2026 reports thinning of more than half a metre a year over recent decades, and a retreat rate that has accelerated from about 0.5 per cent a year across the previous two centuries to between 1 and 2.3 per cent a year over the past 16 years. The Intergovernmental Panel on Climate Change concluded in 2021 that human influence is very likely the main driver of the global retreat of glaciers since the 1990s.
The consequences are mostly gradual: changes in the volume and timing of river flow, and the exposure of new ground and hollows where water can collect. Debris-covered glaciers, common in the Himalaya, can melt differently from clean ice, which is one reason glacier behaviour is not uniform across the range.
Snow: fast clocks, regional contrasts
Seasonal snow works on a much shorter cycle, building each winter and vanishing by summer. Its depth, cover and timing govern when rivers rise, how much water reaches glaciers, and how the ground beneath is insulated. Snow also behaves less uniformly than ice. In the Ecography study of six Himalayan regions from 1999 to 2022, winter snow depth was higher and more variable in the west than the east, and its trends were small and statistically insignificant in every region.
Warming also alters what falls from the sky. The Langtang analysis reports that the freezing level has risen by roughly 100 metres a decade in the monsoon and post-monsoon seasons, and that more of the precipitation at high elevation now arrives as rain rather than snow.
Permafrost: the hidden variable
Permafrost is ground that stays frozen for at least two years. In steep mountain terrain, ice fills the cracks in fractured rock and binds it together. As warming reaches depth, and as meltwater seeps into fissures, that cohesion weakens and water pressure builds within the rock. The result can be a slope that fails without any visible warning, with no glacier at its base to blame. Permafrost is the hardest of the four to observe because it lies beneath the surface, and its role in individual events is usually inferred rather than measured.
In the analysis of the 26 August 2026 collapse at Langtang Lirung, permafrost water is listed among the sources of water in the resulting flood, alongside ice melted by friction, buried glacier ice and displaced river water.
Lakes: where meltwater becomes a hazard
Glacial lakes form where meltwater collects behind moraines, ice or bedrock. They are the most visible link between ice loss and downstream danger, because a lake that breaches or overtops its dam, or is displaced by an avalanche or landslide, can release a glacial lake outburst flood. A peer-reviewed inventory of the China–Nepal Himalaya published in 2026 counts 2,377 glacial lakes whose combined area grew by 27 per cent between 1992 and 2022, and classes 76 as potentially dangerous.
Lakes, however, are only one route to disaster. ICIMOD’s assessment of the August 2026 event concludes that it began with a rock-ice avalanche and not with a lake outburst.
How each is measured
Each process is observed in a different way, which shapes what is known about it. Glaciers are tracked from space and on the ground: satellites measure changes in area and surface height, while field teams measure mass balance directly on a small number of benchmark glaciers. Snow is mapped by satellite and modelled through reanalysis products such as the ERA5 data used in the Ecography study, whose grid of roughly 30 kilometres is coarse for mountain terrain. Lakes are mapped from optical imagery, with the caveat that cloud and seasonal change affect what can be seen. Permafrost is the hard case, because it can be measured directly only in boreholes and is inferred elsewhere from temperature and terrain.
Why the distinction matters
Separating the four processes shows why the same warming produces different risks in different places. Glacier loss changes water supply over decades. Snow shifts it from season to season. Permafrost degradation alters the stability of rock walls in ways that are difficult to see. Lake growth creates hazards that can be mapped, ranked and, in principle, monitored.
The four also interact. Glacier thinning can create the hollows in which lakes form. A higher freezing level exposes more slope to thaw at depth. Changes in snow cover alter how much heat reaches the ground beneath it. Interactions of this kind are why hazard specialists speak of cascades, chains in which a change in one component sets the next in motion.
It also shows that a single event can involve several processes at once. The Langtang Lirung collapse, as described by the analysis, involved bedrock, permafrost and glacier ice together. Treating those as one melting Himalaya obscures where monitoring is strong, as with many lakes and glaciers, and where it is thin.
–Sivarama Murthy Yellamraju
