Every Svalbard Glacier River Sampled Leaks Fossil Methane. Its 1,400 Land-Terminating Glaciers Equal 3,000 Cows, at Most.

Up to 425 times what the atmosphere would put there. That is how much methane a team led by Gabrielle Kleber of the iC3 Polar Research Hub in Tromsø measured in the rivers draining central Svalbard's glaciers, against a baseline of about four nanomoles per liter for near-freezing water in equilibrium with today's air. All 19 rivers they sampled were supersaturated.

Published this week in Nature Communications, the study is the region's largest glacier-meltwater methane survey: 148 water samples from 19 valley glaciers, plus ground-penetrating radar to map which glacier beds are frozen to the rock and which are thawed. Carbon isotopes, together with the ethane and propane traveling with the methane, say the gas is largely thermogenic; it was cooked out of organic-rich shale over geologic time, unlike the microbial methane found under Greenland's ice. Where thawed beds sit on shale, the rivers run rich; where the bed is frozen, or the rock is wrong, there is little to find.

Coverage reached for "feedback loop," and the mechanism supports it: warming thaws more glacier beds, thawed beds over shale leak more fossil methane, and methane warms the planet. The coverage never carried forward the group's own estimate of how much gas that is, so we did, working from the abstract, editor's summary, and press reports, then put it next to the global budget. What came out is a rounding error with an interesting mechanism attached.

Water	Dissolved methane	Times atmospheric equilibrium
Near-freezing river water in equilibrium with air	~4 nM	1x
Peak of 19 glacier rivers, new study (2026)	~1,700 nM	425x
Peak of Vallåkrabreen river, summer 2021	3,170 nM	~800x
Groundwater springs at glacier fronts, late winter 2021 and 2022	not comparable	up to 600,000x

Putting the Flux Next to Something

A 2023 Nature Geoscience paper estimated that groundwater springs in front of retreating glaciers could emit more than 2,000 tonnes of methane a year across Svalbard. A 2025 Biogeosciences study of one 20-square-kilometer glacier, Vallåkrabreen, measured about a tonne leaving the catchment in a melt season, two-thirds of it flushed out by the river, then scaled that to 182 to 368 tonnes a year for the archipelago's roughly 1,400 land-terminating glaciers. The new paper explains why the number swings so widely from glacier to glacier; as far as its public summaries show, it does not revise the total.

Put the top of that range next to something. Saunois and colleagues' Global Methane Budget puts total emissions at about 575 million tonnes a year, so Svalbard's glacier rivers, at 368 tonnes, are one part in 1.6 million. A dairy cow, counting enteric fermentation and manure, emits about 117 kilograms a year, which makes the entire archipelago's meltwater flux a herd of 1,600 to 3,100 cows. Cow methane recycles carbon already in circulation while shale methane adds carbon buried for a hundred million years, which is why the IPCC's fossil-methane factors run slightly higher; even so, at the 20-year value of 82.5 the rivers come to 15,000 to 30,000 tonnes of CO2-equivalent a year, and at the 100-year value of 29.8 they come to 5,000 to 11,000 tonnes, the exhaust of 1,200 to 2,400 American cars.

Methane source	Tonnes CH4 per year	Dairy-cow equivalents	Share of global emissions
Svalbard glacier rivers, land-terminating (2025 scaling)	182 to 368	1,600 to 3,100	1 in 3.2 million to 1 in 1.6 million
Svalbard glacial groundwater springs (2023 estimate)	2,000+	17,000+	about 1 in 290,000
Norway offshore oil and gas (2024)	10,871	93,000	1 in 53,000
World total (2010 to 2019 mean)	575,000,000	about 3x the world's 1.5 billion cattle	1 in 1

Vintages differ, so treat the ratios as order-of-magnitude.

A Cross-Check That Comes Out Smaller

That 182-to-368 range came from multiplying one rich glacier's season by a glacier count. Try it the other way, water times concentration. Svalbard's glaciers shed about 41 gigatonnes of meltwater a year, land-terminating glaciers hold 29 percent of the ice area, and at the average melt rate that gives them roughly 12 gigatonnes, so call it 10 to 20. Reaching 368 tonnes from 15 gigatonnes needs a summer-long average of about 1,500 nanomoles per liter, discharge-weighted across all of them. Vallåkrabreen can do that; back-calculating from the river's 0.65 tonnes, 20 square kilometers, and a typical 1.2 meters of melt gives a mean near 1,700, about half its 3,170 peak. But in the new survey of 19 rivers, the highest single reading anywhere was about 1,700, and a mean cannot sit at 90 percent of the maximum.

Neither paper reports a discharge-weighted mean, so treat what follows as a sensitivity test rather than a measurement: if the typical river averages 100 nanomoles, the flux is 16 to 32 tonnes; at 400, it is 64 to 128. A range built on the richest glacier measured so far looks like an upper bound for the archipelago.

Per square kilometer, Vallåkrabreen's river moved about 0.033 tonnes of methane per square kilometer of glacier per year; Greenland's Leverett Glacier moved 2.8 to 6.3 tonnes from a 600-square-kilometer catchment, or 0.005 to 0.010. On the only two catchments anyone has measured this way, the shale-bedded Svalbard glacier is three to seven times more productive than Greenland's microbial plumbing. Geology sets the rate.

Two Things No One Has Measured

Every measurement here comes from glaciers that end on land. Svalbard's 186 tidewater glaciers end in fjords, and they cover 23,986 square kilometers, 71 percent of the archipelago's ice. At the land-terminating rate they would add 450 to 900 tonnes a year, released under the sea surface where bacteria would eat much of it before it reached air, and since tidewater ice sits mostly off the shale basin and calves rather than melts, that is a strong upper bound.

Melt is the other open variable. Summer 2024 stripped 61.7 gigatonnes of ice off Svalbard, one percent of everything it has, and pushed runoff to 72 gigatonnes, more than twice the 1991 to 2020 average the PNAS team used, so a 2024-type season, scaled against the 41-gigatonne meltwater baseline, would carry 320 to 650 tonnes, though Vallåkrabreen's flushing pulses suggest a bigger flood dilutes as much as it mobilizes. The paper's own word for the future is "initially." Land-terminating valley glaciers are the thin ones, and thin ice does not stay ice.

What You Can Do

Ask for the denominator. "425 times atmospheric" is a concentration, and a concentration alone cannot tell you whether a source matters; the two-line test in this article, tonnes divided by the global budget and tonnes divided by 117 kilograms per cow, works on any new methane source and takes a minute.

For anyone modeling Arctic carbon, the useful output of this paper is a map. Thawed bed plus organic-rich shale equals a leak, and both layers already exist as datasets for the Canadian Arctic Archipelago, Alaska, and the Greenland margin. Overlay them and you know where to send the next 148 bottles.

Methane-abatement money should stay where the tonnes are. Norway's offshore industry emitted 30 to 60 times more methane than Svalbard's rivers in 2024, and satellites have caught the global oil and gas industry leaking half again what it reports. Fix the pipe you can reach.

Limitations

We worked from the Nature Communications paper's abstract, editor's summary, and press coverage rather than its full text, so any Svalbard-wide flux inside the full text is not reflected here. Summer river sampling misses winter spring emissions and gas that bubbles out directly rather than dissolving, and dissolved methane exported by a river is not the same as methane reaching the air, since some oxidizes on the way; the first two biases point up, the third points down, and none of the ranges here is a confidence interval. Our cross-check assumes land-terminating glaciers account for 10 to 20 gigatonnes of runoff, a split the runoff model does not publish, and the Vallåkrabreen back-calculation assumes 1.2 meters of melt, the archipelago average. Bundling enteric and manure emissions in the cow figure flatters the glaciers. The Greenland comparison is two points from different years, and Leverett's figure was a lower bound. Our four-nanomole equilibrium is back-calculated from the group's statement that 3,170 nanomoles was "nearly 800 times" equilibrium.

The Strongest Case Against This Analysis

Smallness is the wrong test for a new source, and the history of the methane budget proves it. Every underestimate started with a first measurement that looked negligible: oil and gas leaks, reservoir surfaces, thermokarst lakes. This paper's finding is a mechanism, and the same ice sits across the Canadian Arctic and Alaska, almost none of it sampled for methane. Springs the same group measured in winter run 600,000 times equilibrium and already amount to a fifth of what Norway's offshore industry reported for 2024 (the 2023 paper said a tenth, against a larger, older Norway figure). Seventy-one percent of Svalbard's ice drains where no one has looked. And the deeper version of the objection is about the reservoir rather than the rate: the shale under Svalbard holds gas accumulated over a hundred million years, uncapping it is a centuries-long process, and this year's leak rate says little about the integral. Summing 19 rivers and declaring the Arctic safe is precisely the error that made the oil-industry satellite numbers a surprise.

That argument is right about the mechanism and wrong about what it would take to matter. Take Svalbard's rivers and springs together, about 2,400 tonnes, multiply by ten to stand in for every shale-bedded glacier outside the Greenland Ice Sheet, and you reach 0.004 percent of global emissions; run Greenland's entire 1.7 million square kilometers at Leverett's microbial rate and you add 8,500 to 17,000 tonnes, one part in 34,000 to 68,000. And because methane lasts about twelve years in the atmosphere, its warming tracks the rate you just measured, not the size of the reservoir. This mechanism deserves a research program. The press framing does not yet deserve the word "loop."

The Bottom Line

Svalbard's glaciers are leaking fossil methane, the leak is controlled by whether the bed is thawed and whether the rock is shale, and warming thaws more beds; all of it is true and newly well documented. It is also, at every scale anyone has measured, a source you could offset by retiring one mid-sized dairy farm. Keep watching anyway, because the Arctic has a lot of shale, a lot of ice on top of it, and a warming rate that is turning frozen beds into thawed ones faster than anyone has sampled them.