Estimated stocks of circumpolar permafrost carbon with quantified uncertainty ranges and identified data gaps

Picture the ground beneath your feet as a giant freezer. Not a neat one, but crammed with ancient plant bits and dark, fibrous muck — the remains of millennia of life and death in cold places. That's soil organic carbon, and in the far north, a lot of it is locked into permafrost. Warm that freezer, and some of that carbon can move, decompose, and leak into the air as greenhouse gases. For years, we've known the stakes but not the size of the stash. The maps were patchy, the deep layers were guesswork, and the High Arctic was almost a blank. So, Hugelius and colleagues set out to redraw the ledger, this time with uncertainty bounds you can actually use. Here's the cleanest way to hold the result in your head. Across the northern circumpolar region, they estimate about 1,307 billion tonnes of carbon — 1,307 petagrams — stored in soils and deep sediments, and they bracket that with a 95 percent uncertainty range of 1,140 to 1,476. Most of it sits in permafrost terrain, about 999 petagrams. Of that, roughly 822 petagrams are still perennially frozen. So, the freezer is big. But crucially, a smaller slice than we thought is in the hard-frozen category, which matters for how fast it could thaw and move. Now, that headline sits on top of a depth story. If you only look down three meters, you're capturing 1,035 petagrams. Nearly half of that is in the top meter, about 472 petagrams, and almost half of that again is in the very top 30 centimeters, roughly 217 petagrams. That's the living layer, the one that thaws each summer, breathes with microbes, and responds quickly to climate. Deeper down, there's still a lot: about 827 petagrams in the top two meters and another 207 between two and three meters. The point is simple. The surface matters first for fluxes, but the deep pool is not trivial, especially once you start thinking in decades instead of years. To build these numbers, they didn't just average a pile of soil samples. They followed a three-step upscaling playbook that Hugelius laid out a decade earlier. First, calculate carbon storage for each individual soil profile — a pedon — at standard depths. Second, sort those pedons into meaningful classes and compute average storage and variability for each class. Third, multiply those class means by mapped areas to roll up to regions and the whole circumpolar belt. The secret sauce here is the map: the Northern Circumpolar Soil Carbon Database, version 2. It doesn't just draw continents; it splits the landscape by physiography — places with thick sediment overburden versus thin, rocky terrain — and treats the High Arctic separately because soils there are young and shallow. Those classes matter. Permafrost-affected soils — Gelisols — are subdivided into Turbels, Histels, and Orthels; organic-rich Histosols get their own lane. For the top meter, they leaned on a large pedon set — 1,778 profiles — and for the 1 to 3 meter depths, they assembled a new database with 518 profiles down to two meters and 351 to three. Then they stress-tested consistency. Comparing independent datasets for the 0 to 1 meter layer, they found no significant differences for Orthels and Histels, but Turbels and non-permafrost mineral soils tended to look richer in the deeper-focused dataset, while Histosols looked leaner. That's not just bean counting. It tells you where sampling strategy nudges estimates and where it doesn't. Physiography also frames the uncertainty. Lowland regions with thick sediment — think broad Arctic plains — are better constrained. You tend to find higher carbon densities there, especially in organic soils, and the class averages are backed by more profiles. Mountains and uplands with thin overburden are a different beast. There, Turbels drive much of the error bar, and replication is scarce. The High Arctic is the starkest case. It only covers about six percent of the area, but still packs roughly 34 petagrams into its top three meters. Most of that is near the surface — about 24 petagrams in the top meter, and a big chunk of that, around 10 petagrams, in the top 30 centimeters. It's a small place with a surprisingly consequential top layer. But three meters is an arbitrary line in the ground. Below that, two kinds of deposits dominate the north's deep carbon: deltaic alluvium and the Yedoma domain. Both are geological filing cabinets for old organic matter, layered with ice, and they're hard to measure. Hugelius and colleagues tackle them separately, and with extra care on depth. Start with the deltas. The northern permafrost belt has a dozen major ones, covering about 75,800 square kilometers. If you strip off the upper three meters to avoid double counting, and you also pull out open water and massive ground ice, you end up with a subaerial alluvium volume below three meters of roughly 3,500 cubic kilometers. That's a lot of wet sand and silt. It averages about 54 meters thick. The team then combines those volumes with literature-based carbon densities to get a stock. The answer they land on is about 91 petagrams of carbon below three meters, with a wide uncertainty of plus or minus 52. A big fraction of that is frozen — on average around 84 percent by area, translating to roughly 69 petagrams in perennially frozen alluvium. The Mackenzie and Lena deltas loom largest in that ledger, at around 34 and 23 petagrams respectively, with the Yana contributing about seven and the rest of the deltas sharing the remainder. Now to Yedoma, which is its own Arctic story. These are ice-rich silts from the late Pleistocene that blanket huge swaths of Siberia and Alaska. The Yedoma region spans on the order of 1.39 million square kilometers, but only about 30 percent of that — roughly 416,000 square kilometers — is intact Yedoma today. More than half has been churned by thaw and refreeze into thermokarst deposits. Below three meters, this domain holds a striking amount of carbon: about 181 petagrams, plus or minus 54. Of that, roughly 74 petagrams are in the still-intact Yedoma, and the rest — a little over a hundred — is locked in refrozen thermokarst. If you zoom out to all permafrost in Yedoma terrain, not just the deep slice, the total is around 213 petagrams, with about 83 in intact Yedoma and roughly 130 in thermokarst. That tells you most of Yedoma's frozen carbon today sits in the reworked, lumpy terrain left behind by past thaw. The way they get there is also worth a beat. For Yedoma, they draw on profiles compiled by Marcus Strauss and colleagues — 22 intact Yedoma and 10 thermokarst sections — and then they bootstrap. Ten thousand times, they resample observed thickness, organic carbon content, bulk density including segregated ice, and the share of wedge ice, to build a distribution of plausible totals. The mean is the estimate, and the fifth and ninety-fifth percentiles set the bounds. For the deltas, they extend the classic approach by Eduard Tarnocai. They update mapped delta extents, assign mean alluvium and lake depths, adjust for permafrost extent and segregated ice, and then, crucially, shave off the top three meters so those deep numbers don't overlap with the soil map. All of this sits inside a clear accounting of uncertainty. The soil upscaling uses 95 percent confidence intervals based on within-class variability and the area or volume that each class covers, and those uncertainties are combined with a formula that accounts for covarying pieces rather than treating everything as independent. They also separate two kinds of error we tend to conflate: representation error, when you just don't have many samples in a class or region; and spatial error, which is what happens when the map itself is off. The reported confidence intervals include the first, where possible, but not all of the second. And that matters. The team notes that if you assumed even a 10 percent spatial error in the mapping, you could widen the total uncertainty on the 1,307 petagrams by something like 30 to 200. The deltas underline the point: there are only a handful of depth observations — one from the Lena, five from the Mackenzie — and in their budget, depth uncertainty alone contributes about 49 petagrams of the error bar, while uncertainty in carbon density adds about 18. It's also a methodological fork in the road compared to some earlier syntheses. A continent-based upscaling in work led by Jennifer Harden and Gustaf Hugelius in 2013 didn't split by physiography, and it produced much higher Gelisol totals — around 1,060 petagrams — than the 727 petagrams for Gelisols implied here. That single decision, to treat thick sediment lowlands and thin, rocky uplands as distinct, pulls the central estimates and the confidence intervals in a different direction. Compared with Tarnocai's 2009 circumpolar tally, Hugelius and colleagues pull several components down: the 0 to 3 meter soil total is smaller, the deep deltaic and Yedoma stocks come in lower once overlap and updated extents are handled, and the perennially frozen slice of the pie shrinks substantially to about 822 petagrams. What does that mean outside a spreadsheet? First, this is a more realistic baseline for the permafrost-carbon feedback. The surface pool is big and active, especially in thick-sediment lowlands, which is exactly where near-term warming is likeliest to wake microbes and move carbon. But the deep pools, particularly in deltas and Yedoma, are not afterthoughts. They're slower to thaw, often protected by ground ice, and hugely consequential on multi-decade horizons. Second, the error bars point like signposts. If you want to cut uncertainty in half, don't drill randomly. Go to the High Arctic, where soils are poorly developed but matter for surface fluxes. Sample Turbels in thin-overburden terrain, which are driving noise in the 1 to 3 meter band. Get serious about deep coring in a few key deltas and across the Yedoma domain, because right now a half-dozen depth observations can sway a continental budget. There's also a quiet but important modeling implication. Earth system models need boundary conditions. With this map, modelers can assign carbon densities by physiographic class, layer them by depth, and turn the dials on permafrost extent with some confidence in how uncertainties propagate. The authors are frank that their bounds are conservative — lab measurement error and all map-based inaccuracies aren't fully captured — but conservative doesn't mean vague. It means we have a floor under what we know and a clear plan for raising the ceiling. If you're listening for a single takeaway to carry into a policy meeting or a classroom, here it is. The north's freezer holds around 1.3 trillion tonnes of organic carbon, and roughly 800 billion tonnes of that are still locked hard in permafrost. The contributions are layered — thick lowlands heavy with near-surface carbon, deltas and Yedoma stockpiling the deep stuff — and the biggest blind spots are where the ground is thin or the data are thinner. Close those gaps, and we'll not only sharpen the numbers. We'll get a much cleaner read on how a warming Arctic talks back to the climate.