Connect with us

NEWS

Lake-Ending Glaciers Drive Half of Global Ice Loss

Lake-ending glaciers lose ice 57% faster than land fronts, a global census finds, so spreading glacial lakes can raise future melt rates.

Published

on

Lake-terminating glaciers and other water-ending ice tongues are 3.2% of the global count, yet they accounted for more than half of glacier ice loss from 2000 to 2019.

Joseph Mallalieu, a glaciologist then at the University of Birmingham and now at the University of Leeds, led a globally complete classification of glacier termini published on 23 October 2026. After climate, size and slope were held constant, a lake at the front was still tied to specific ice loss about 57% higher than on land.

Just 3 Percent of Glaciers Drain Half the Ice

The team tagged every glacier in the Randolph Glacier Inventory version 6.0, 216,502 ice bodies excluding the Greenland and Antarctic ice sheets. Land fronts dominate the headcount. They do not dominate the ice.

Land-ending glaciers are 209,563 of the total, or 96.8%. Lake fronts number 3,548 (1.6%), marine fronts 3,222 (1.5%), and ice-shelf fronts 169 (0.1%). Add the three water classes and you get 6,939 glaciers, 3.2% of the inventory.

Those water-ending glaciers hold 401,291 of 746,093 square kilometres of glacier area, 53.8%. Marine fronts alone drain 231,383 km² (31.0%), with lakes at 85,950 km² (11.5%) and shelves at 83,958 km² (11.3%). Land still has the single largest slice of area at 344,802 km², 46.2%, because the typical land glacier is small.

ICE LOSS BY TERMINUS TYPE, 2000 TO 2019

Terminus Glaciers Share of count Area (km²) Share of area Mass change (Gt a year) Share of loss
Land 209,563 96.8% 344,802 46.2% -131.2 49.4%
Lake 3,548 1.6% 85,950 11.5% -63.8 24.0%
Marine 3,222 1.5% 231,383 31.0% -60.4 22.8%
Shelf 169 0.1% 83,958 11.3% -10.2 3.9%
Water-ending total 6,939 3.2% 401,291 53.8% -134.4 50.6%

The mass numbers come from the Hugonnet et al. elevation-change record that Mallalieu’s group split by terminus. That parent study put global glacier loss, ice sheets aside, at 267 gigatonnes of ice a year from 2000 to 2019, equal to 0.74 millimetres of sea-level rise a year and 21 ± 3% of the observed rise. This 2026 split attributes 265.6 Gt a year across the four front types; the Caucasus and Middle East outlines had no matching mass data, so the two totals are not the same product. Water-ending fronts account for 134.4 Gt a year, 50.6% of the attributed loss. Lake fronts, a sliver of the count, take 24.0% of that ice by themselves.

Lake Fronts Lose Ice Twice as Fast

Raw speeds already sort the four classes. Median surface velocity in 2017 to 2018, from maps by Romain Millan and colleagues, was 5.1 metres a year on land, 7.4 on lakes, 9.7 on marine fronts, and 7.2 on shelves. Speed data existed for 181,603 glaciers, 83.9% of the inventory.

Specific mass change, ice loss divided by area, tells a different ranking. Median rates for 2000 to 2019 were -0.490 metres of water equivalent a year at lake fronts, -0.245 on land, -0.239 at marine fronts, and -0.051 at shelves. The lake median is roughly double the land figure. That pattern held in 15 of the 18 regions with mass data. Coverage here was 203,372 glaciers, 93.9%.

Size and climate could have explained the whole gap. Lake and marine glaciers tend to be larger, thicker, lower and less steep, with lake fronts sitting between land and sea on those measures. The team fitted linear mixed-effects models that held climate, thickness, slope and elevation constant, with a random intercept for each inventory region.

MEDIAN DYNAMICS BY FRONT TYPE

Terminus Median speed (m a year) Median specific mass change (m w.e. a year) Controlled flow vs land Controlled extra loss vs land
Land 5.1 -0.245 reference reference
Lake 7.4 -0.490 +10% -0.162 m w.e. a year (57% greater)
Marine 9.7 -0.239 +44% +0.019 (less loss)
Shelf 7.2 -0.051 not modelled +0.100 (less loss)

After those controls, lake glaciers still flowed 10% faster than land glaciers, and marine glaciers 44% faster. Specific mass loss at lake fronts remained 57% greater than at land fronts, a gap of 0.162 metres of water equivalent a year. Marine and shelf fronts lost slightly less than land once geometry was in the model, by 0.019 and 0.100 metres of water equivalent a year. Shelf speed could not be modelled; only four of those glaciers had usable velocity pixels.

Land and marine glaciers look almost identical on raw specific loss (-0.245 against -0.239) even though they sit at opposite ends of the size range. The paper’s reading is that efficient calving at a tidewater front lets a marine glacier adjust its shape faster, so the area-normalised thinning stays moderate. A lake front does not get that same rapid reset, and the extra thinning shows up in the specific rate.

Why a Lake Speeds a Glacier Up

Water at a glacier snout does work that air cannot. The paper groups the lake effect into a short list of thermo-mechanical hits, each already seen in the Himalaya, New Zealand and Greenland, and now visible in the global medians.

HOW A TERMINUS LAKE ATTACKS THE ICE

  • Calving: The front can shed icebergs once it stands in water, so ice leaves as blocks instead of only as melt.
  • Underwater melt: Lake water delivers heat to ice below the waterline and can undercut the cliff, which then fails.
  • Buoyancy: Water can lift and float the lowest ice, which favours more calving.
  • Basal lubrication: Lake water that works under the front cuts friction, so the ice above can run faster.

Those four processes also steepen the lower glacier, which raises the driving stress. Lakes are small compared with the ocean and mix slowly, so they can store summer heat against the ice. The authors treated every glacier that touched a terminus lake, even a modest one, as lake-ending. Glaciers with only light lake contact should behave more like land ice, which means the reported lake-versus-land gap is likely a conservative figure.

Glaciers ending in lakes really stand out: they’re losing ice particularly quickly. We don’t yet know exactly why. It could be the lakes themselves, the size and shape of the glaciers, or a bit of both. Untangling that is the next big question, especially as glacier lakes are becoming more common in a warming climate.

Joseph Mallalieu, glaciologist, University of Birmingham newsroom

Ice lost below the waterline is easy to miss from space, which is one reason a lake front can look even hungrier once you go looking for it. A 2023 Himalayan study found that accounting for subaqueous melt raised estimated mass loss at lake-terminating glaciers in that range by about 7%, with a larger miss in the central Himalaya where lakes had grown fastest. Mallalieu’s global medians sit on surface-elevation change, so they already capture thinning on the ice; they still cannot see every block that calves and melts out of view.

Iceland Sends Two-Thirds of Its Ice Into Lakes

The imbalance is not evenly spread. In inventory regions 11 to 16, the lower-latitude belts that include the Alps, the Caucasus, Central Asia and the low-latitude tropics, more than 87% of glacier area still drains to land. At high latitude the map flips.

Iceland sends 65.1% of its glacier area into lakes. The Southern Andes send 37.8%. Alaska, Arctic Canada South, Scandinavia and New Zealand each send more than 21% of area to lakes. Lake fronts are most numerous around the Greenland Periphery (754) and also exceed 300 in Western Canada and USA, Arctic Canada South, Central Asia, South Asia East and the Southern Andes. Shares of 7 to 10% of glacier numbers show up in the Russian Arctic, Scandinavia and Iceland.

Marine ice is a high-latitude story. Those fronts exist in 9 of 19 regions. They are 48.7% of termini in the Antarctic and Subantarctic (1,339 glaciers) and 35.1% in the Russian Arctic, with another 903 around the Greenland Periphery. Marine ice drains 63.2% of area in the Russian Arctic, about 45% in Arctic Canada North, the Greenland Periphery and Svalbard, and 36.2% in the Antarctic and Subantarctic. Ice shelves appear only in the Russian Arctic and the Antarctic and Subantarctic, where they take 1.6% and 62.6% of area.

Non-land fronts already dominate regional mass loss in Iceland, the Southern Andes, Svalbard and the Russian Arctic, each with more than 50% of losses at lake or marine glaciers. Lake-ending ice took more than 20% of mass loss in 7 of 19 regions. Marine-ending ice took more than 30% in 5 of 19.

A Shift From Sea to Land Cuts Speed

The census is a snapshot, and the snapshot is already moving. Marine glaciers are pulling back onto rock around the Arctic. Land glaciers are growing lakes at their snouts almost everywhere else. The two switches do not cancel.

Mallalieu’s models imply a marine glacier that retreats onto land should slow, because the 44% flow premium is tied to the water. A land glacier that gains a terminus lake should speed up only modestly, about 10%, but its specific mass loss can jump by that 57% margin. If glacial lakes keep spreading, the global loss rate can rise even if the climate forcing stays on the same path.

The lake boom is not a forecast. Dan Shugar and colleagues, working with more than 250,000 Landsat scenes, found that the number of glacial lakes rose from 9,414 in 1990 to 1999 to 14,394 in 2015 to 2018, a 53% increase. Area rose 51%. Estimated volume rose about 48%, to 156.5 km³, about 0.43 millimetres of sea-level equivalent. NASA’s briefing on that survey put the same growth in plain terms: glacial lake volume rose about 50 percent since 1990, to roughly one-third the volume of Lake Erie. Mallalieu used Shugar’s multi-year mosaics for 2000 to 2018 to decide which glaciers touched a terminus lake.

On the marine side, Jan Kavan and colleagues mapped 2,466 km of new Northern Hemisphere coastline exposed by retreating tidewater glaciers between 2000 and 2020, 123 km a year, with 66% of that length in Greenland. Each kilometre of new coast is a glacier that has left the sea. Under Mallalieu’s velocity result, that is also a glacier that should shed its marine speed premium.

THE SWITCHING CLOCK BEHIND THE CENSUS

  1. 1990 to 1999: Shugar’s first epoch counts 9,414 glacial lakes larger than 0.05 km².
  2. 2000 to 2010: Most Randolph Glacier Inventory 6.0 outlines are drawn, the base map for this census.
  3. 2000 to 2018: Four Shugar mosaics mark maximum lake extent; any glacier within 60 m of a terminus lake is tagged lake-ending.
  4. 2000 to 2019: Hugonnet elevation change supplies the mass-loss rates split by front type.
  5. 2000 to 2020: Kavan’s group measures 2,466 km of new Arctic and sub-Arctic coast as marine glaciers pull onto land.
  6. 2015 to 2018: Glacial lakes number 14,394, with volume near 156.5 km³.
  7. 2017 to 2018: Millan velocity and thickness maps, plus Greenland MEaSUREs mosaics, give the speed medians.
  8. 23 October 2026: Nature Communications publishes the four-class global tally.

The two transitions will often happen in different places. New lakes are spreading through mountain ranges that already run on meltwater for farms and cities. New rock coasts are opening in Greenland, the Canadian Arctic, the Russian Arctic and Svalbard, where the ice that is slowing is also the ice that had been feeding the sea most directly.

Ice Models Still Skip the Lake Terminus

Global glacier models have improved at putting ice flow into the forecast. Many still treat a lake front as if it were land, or they fold it into a simplified calving rule built for tidewater ice. David Rounce and colleagues, in a 2023 projection for every glacier outside the ice sheets, did add iceberg calving at ocean fronts and the effect of surface debris. Their range for 2100 was a 26 ± 6% mass loss if warming is held to 1.5°C, and 41 ± 11% at 4°C, relative to 2015. Lake-front physics was not in that toolkit, and Mallalieu’s paper says that omission is still the rule.

That gap matters most where non-land ice already dominates the regional budget: Iceland, the Southern Andes, Svalbard, the Russian Arctic, and the lake belts of Alaska and New Zealand. More than half of global glacier area, and more than half of recent mass loss, sits on glaciers that can respond in a non-linear way when the front changes. A model that only melts the surface will miss the extra 0.162 metres of water equivalent a year that the lake term carries in this census, and it will miss the slowdown when a tidewater glacier becomes a land glacier.

The last decade has seen a huge increase in the global glacier and climate data available to researchers. Bringing these datasets together lets us compare how glaciers in different settings are behaving. As more glaciers shift from one setting to another, understanding those differences will only become more important.

Nicholas Barrand, Associate Professor of Glaciology, University of Birmingham newsroom

The classification itself is now public. The glacier terminus inventory on Zenodo ships with the climate and geometry fields the models need, and the processing code sits on GitHub. Three Southern Andes glaciers that drain into Lago Argentino and Lago San Martín were added by hand because those lakes exceeded Shugar’s 200 km² mapping cap. Everywhere else the rules were automatic: a 60 m buffer around the outline, a 0.05 km² minimum lake, and an elevation filter that threw out ice-marginal ponds sitting more than 100 m above the lowest ice.

People who build those models have been asking for lake frontal ablation for years, including on New Zealand’s expanding proglacial lakes. The 2026 census does not give them a full lake-physics scheme. It does give them a map of which glaciers need one, and a measured 57% specific-loss premium to beat.

Frequently Asked Questions

What Counts as a Lake-Terminating Glacier?

A glacier counted as lake-ending if its Randolph Glacier Inventory 6.0 outline came within 60 metres of a terminus lake of at least 0.05 km² at any point from 2000 to 2018, using Shugar’s dissolved multi-year lake mosaics and a 100 m elevation filter to drop ice-marginal and supraglacial ponds. Svalbard used a 2008 to 2012 aerial lake map with the same filters. The Antarctic and Subantarctic lake tag was taken from the inventory as published. Any glacier already classed as marine kept that label even if a lake also touched it.

Why Did the Team Use Randolph Glacier Inventory 6.0?

Version 7.0 was already out, with changes to 73% of outlines and a 27% rise in glacier count but only a 0.1% rise in total area. The Hugonnet mass-change grids and the Millan velocity and thickness maps were built on 6.0 outlines, so the authors kept 6.0 to avoid mixing polygons. Compared with a later lake inventory by Steiner and colleagues for version 7.0, Mallalieu’s global lake shares sit close to Steiner’s stricter classes, the ones judged likely to affect ice flow, and well below Steiner’s widest net, which used a 0.01 km² lake cutoff instead of 0.05 km².

Are the Greenland and Antarctic Ice Sheets Included?

No. The census covers mountain glaciers and ice caps in the Randolph Glacier Inventory, including ice around the Greenland Periphery and in the Antarctic and Subantarctic, but not the two ice sheets. For the statistical models the authors also dropped 461 glaciers with estimated outlines, 955 peripheral Greenland ice bodies strongly tied to the ice sheet, 1,343 surge-type glaciers, and 2,729 glaciers within 5 km of a volcano, so the speed and specific-loss tests would not be pulled by surges or geothermal heat.

What Is Specific Mass Change?

Specific mass change is ice loss divided by glacier area, reported in metres of water equivalent a year, so a small alpine glacier can be compared with a huge tidewater outlet. A lake glacier with a median of -0.490 metres of water equivalent a year is thinning, in that area-normalised sense, about twice as fast as the land median of -0.245, even when the marine glacier next door dumps more total ice because it is simply bigger. Total mass change in gigatonnes is the figure that feeds sea-level rise; the specific rate is the one that shows whether the front type itself is hungry.

Harry is the editor of COVER 365, an independent publication he owns and runs, and a journalist of ten years who moved from reporting into editing. Anything the site reviews has been used before it is judged. A phone, a car, a game or a piece of travel gear is tested in ordinary conditions, its measured results are set against the maker's specification sheet, and where the two disagree the article says which one to trust and why. No product gets a verdict Harry has not earned by using it. Off the test bench, the same rule of primary evidence applies: business stories come from filings and results, science from the published paper, sports from the governing body's records, and news from statements and transcripts rather than second hand accounts. Coverage runs across technology, auto, gaming, lifestyle and travel as well as news, business, science, sports and entertainment, for readers in every part of the world. Every figure is checked before publication and corrected publicly under a stated policy when wrong. Reader mail is answered at support@cover365.in.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending