An archive whose units look identical and are not — every metre the same length, no two metres the same amount of the past, and ordinary use spends the scarcest metres fastest
R841’s Finnish reserve works because a sack of grain is a sack of grain. The level is kept, the contents flow through and are eaten, and nothing is lost by the substitution — a fresh sack really is the old sack, for every purpose the reserve exists for. Turn to the kept thing where that substitution fails completely: a store whose units are all the same size and shape, priced in the same currency, and utterly non-interchangeable, because each unit is a different stretch of unrepeatable time.
What is kept. An ice core is a column of old weather. Snow falls, is buried, and is squeezed into ice, sealing in tiny bubbles of the air of its year — and those bubbles are, in the words ice-core scientists keep using, the only direct archive of what past atmospheres were actually made of. No instrument can be sent back to sample the air of 100,000 years ago; this is the sample. Cores from Greenland and Antarctica are curated in a handful of freezers. The US repository, the National Science Foundation Ice Core Facility at Lakewood, Colorado, holds roughly 22,000 metres of core at about −35°C and distributes pieces of it to researchers around the world against approved proposals. Analysis, in the ordinary course of the science, is destructive: chemistry is commonly done on melted ice, and the trapped air is freed by crushing or grinding the sample under vacuum. The knowing spends the known.
A metre is not a metre. Layers thin as they are buried — the weight above presses them flat and the deep ice spreads sideways — so the rate at which a core stores time rises steeply with depth. The EPICA core at Dome C in Antarctica runs 3,270 metres and reaches back about 800,000 years: a whole-core average of roughly 245 years per metre, which already hides everything that matters. Summaries of deep polar ice describe less than an inch of it holding a century — over 3,900 years to the metre. And the Beyond EPICA team, drilling for the oldest ice yet, expected before drilling that the bottom metre of their core would hold some 20,000 years, its bubbles pressed out of existence into the crystal lattice itself. Same cylinder, same diameter, same ten centimetres on the saw — and the bill, in years of the Earth’s past, varies by a factor of eighty.
Use concentrates where the stock is thinnest. Finland’s rotation spends the reserve indifferently, oldest first, and indifference is safe when units are fungible. Here the ordinary demand of the science is not indifferent at all. Nobody queues for the thick, young, recent metres, which are abundant and — at the price of another expedition — in principle re-drillable. The requests gather on the transitions: the ends of ice ages, the abrupt warmings, the deepest and oldest bands, exactly the intervals that exist in the least length, compressed the thinnest, and backed by no second copy. Drawing on this archive in ordinary times preserves everything about it that is common — the methods improve, the recent record thickens — and spends, sample by sample, precisely the part no purchase can put back. A replacement core is not an order you place; it is a decade-scale expedition, and for mountain glaciers that have since melted down past the layers in question, it is not even that: the ice that held those years no longer exists anywhere.
The keepers’ answer is rationing, not rotation. Because the stock cannot flow, it is fenced. Processing plans for major cores hold a substantial fraction back untouched — for the WAIS Divide core, about a third or more of each metre-long section was retained for future studies — samples are allocated against proposals rather than taken, and laboratories invest in material-saving methods: one Bern group’s recent technique aims to measure the greenhouse gases in the oldest ice while losing none of the extracted air, so the same air can be measured again. None of this refills anything. It only decides, carefully and in committee, which years the present is allowed to spend — and reserves a share for future scientists whose questions and instruments the present cannot foresee.
The toy. Schematic only. The saw always takes the same 10 cm. What changes is what 10 cm costs, at three compressions anchored as described on this page: the EPICA whole-core average (~245 yr/m), a deep band at a century to the inch (~3,937 yr/m), and the reported expectation for the oldest ice (20,000 yr/m). Rates are site-specific; treat these as three labelled points on a steep curve, not a profile of any one core.
The shelf is full. Every sample below spends real, unrepeatable years.
Three narrowings, so the claim stays honest. First, not every measurement is destructive — imaging and some physical-properties work leave the ice available, so spending is a property of the method, not of curiosity itself. Second, a replicate core is not impossible; it is merely an expedition, and a new core is a sibling of the old one, not the same ice — its layers correlate, they do not coincide. Third, the 20,000-years-to-the-metre figure is a pre-drilling expectation as reported in 2023, not a measured rate from a retrieved core, and the deep band’s inch-a-century is a summary figure, not a constant — real thinning curves are site-specific and the bottommost ice can be disturbed or, where geothermal heat wins, melted away entirely, which is what caps Dome C at its ~800,000 years in the first place.
What is actually kept. Finland keeps a level and lets the contents go. This archive can only keep the contents themselves — a finite run of frozen years, each existing in exactly one place, in a length that shrinks with every ordinary use. Its keepers therefore practise the opposite of rotation: they hold back a third, ration the rest by proposal, and spend the present’s share knowing the exchange rate worsens with depth — ten centimetres here, a human lifetime of climate there, two hundred centuries at the bottom — and that the metres everyone most wants are the ones the metre measure lies about most.