Weighing The World's Ice

AOf the 215,000 glaciers listed in the world inventory, fewer than one in a thousand is ever visited by anyone carrying an instrument. The Global Glacier Register, founded in 1986 to collate national measurements, treats 42 of them as reference glaciers: bodies of ice with unbroken records of at least thirty years. Thirty-one of those 42 lie in the ranges of Europe and North America. The method itself has scarcely altered since the 1940s. Aluminium stakes are drilled several metres into the ice in autumn and read again the following spring; pits dug into the snowpack give its density; the two figures together yield the mass balance, the difference between what a glacier gains in winter and loses in summer, expressed as a depth of water spread evenly across its surface. A single stake season on a remote glacier costs upwards of $40,000 and depends on helicopters. It is not a method that can be scaled to 215,000.

BElise Corbeau, a geodesist at the Grenoble Cryosphere Unit, regards that arithmetic as settled. Since 2018 a laser altimeter in polar orbit has recorded the surface elevation of every glacier larger than half a square kilometre, and repeated passes turn those elevations into a change in volume. A second technique, running since 2002, weighs entire mountain ranges by detecting the minute tug they exert on a pair of satellites. Corbeau's group combined the two and put the global loss between 2000 and 2023 at 267 billion tonnes a year, a figure some 12 per cent higher than the one the field network had produced for the same period. She does not present the satellites as infallible. What they record is volume, not mass, and converting the first into the second requires an assumption about the density of whatever has gone, conventionally 850 kilograms per cubic metre, a value that behaves well when averaged over a decade and badly over a single year.

CRahul Menon, who runs the Karakoram Field Programme, does not dispute the satellite totals. His objection is that a total is the wrong thing to want. A stake network separates the two halves of the balance, how much snow arrived and how much ice went, whereas an elevation change records only the residue after the two have cancelled. Two glaciers losing identical volumes may be doing so for opposite reasons, and only the older method can say which. Menon also lists what orbit cannot see. Where a glacier is buried under rock debris, a layer thicker than about five centimetres insulates the ice and slows melting, while a dusting of two centimetres or less darkens the surface and speeds it up; the elevation signal is the same in both cases. Meltwater that refreezes deep inside the glacier, meanwhile, registers from above as no loss at all, though it has altered the ice profoundly.

DThe contest is less even than the two positions imply, because the ground network has been contracting while the orbital one expanded. The Register's own tables show the number of glaciers submitting field measurements falling from 61 in 1995 to 38 in 2022, as national agencies moved budgets towards remote sensing. Drones and automatic weather stations have replaced some of the lost effort at a fraction of the cost, but they cluster on the same accessible glaciers the stakes did. The Karakoram anomaly remains the standing embarrassment: between 1999 and 2010 several glaciers in that range thickened while almost every other range thinned. The explanation now generally accepted, a rise in winter snowfall driven by a shift in the westerly storm track, emerged only after a decade of argument, and it was reached from weather records and satellite elevation together, neither of which would have been sufficient alone.

EAt a meeting in Zurich in 2024 the disagreement narrowed. Corbeau accepted that the density value her method depends on has been calibrated at fewer than twenty sites, all of them inside the reference network, so the satellite record is not independent of the ground record it is often said to supersede. Menon accepted that the six seasons of stake data he cites come from glaciers of a single type, all fed by avalanche rather than by direct snowfall, and that stake networks say nothing about ice lost where a glacier ends in a lake. What they could not settle was priority. Menon argued that a shrinking calibration base will eventually degrade the satellite product itself, so the ground network should be rebuilt first; Corbeau replied that no realistic budget could restore field measurement to the scale the problem now has, and that money spent on stakes is money withheld from the only instrument that sees everything.

FA third argument has begun to displace both. For the 1.9 billion people living in basins fed by mountain ice, what matters is not the annual total but the timing of the flow. As a glacier shrinks it first yields more water and then less; the turning point, known as peak water, has already been passed in eight of the eleven basins examined in a 2023 assessment, though the authors stressed that for the two largest basins their runoff figures had been modelled rather than gauged. Monitoring designed to weigh ice is poorly suited to forecasting a river, and the instruments that would be needed belong to a different discipline and a different budget. The deeper difficulty is one neither camp has solved. Every trend, from either method, is measured against an inventory of glacier outlines compiled from aerial photographs taken in the 1970s, and that baseline has never been redrawn.