Ice core records

How can bubbles sealed inside ancient ice tell us what the air was like long before anyone measured it?

The claim is remarkable enough to deserve suspicion: crush a fragment of ice drilled from deep in an ice sheet, collect the escaping gas, and you have a sample of the actual atmosphere of a hundred thousand years ago — not a proxy for it, the thing itself. Most evidence about the deep past is inferential; this appears to be direct. So the question is not "how do we read it" but "what exactly is in that bubble, and how old is it?" — and the honest answer to the second part is the surprising one.

Follow a snowflake down. It lands on an ice sheet where nothing ever melts — and that condition is doing real work, because meltwater would dissolve gases and move them between layers, which is why the record comes from the cold interiors of Antarctica and Greenland and not from a temperate glacier. More snow buries it, and the pile compacts into firn: granular, half-way to ice, and crucially still porous. Its pore spaces connect to the surface, and air moves through them freely by diffusion.

How deep does that go? Tens of metres — typically somewhere around fifty to a hundred, depending on temperature and snowfall. Now sit with the consequence, because it upends the naive picture. Throughout that whole column the ice is getting older with depth while the air in its pores is being continually refreshed from above. Only at the bottom of the firn do the pores finally pinch shut and seal. So at the moment of trapping, the ice is already centuries or millennia old, and the air inside it is roughly modern.

That difference has a name — the gas age-ice age difference — and it is not a nuisance detail; it is the central thing to understand about the archive. It ranges from a few decades at wet, fast-accumulating coastal sites to thousands of years at the coldest, driest interior domes, precisely because a slow-accumulating site takes far longer to bury a snowflake through the firn column.

There is a second consequence of the same fact. The pores do not all close at once, and the air below has been mixing by diffusion the whole time, so a bubble does not hold one year's air. It holds a blend, spread over a window that is narrow at high-accumulation sites and wide at low-accumulation ones. That is a genuine limit rather than an inconvenience: an ice core faithfully records how carbon dioxide changed over a century, and at the driest sites simply cannot resolve a single-year spike — the archive has already averaged it away.

Two more corrections are needed before the number means anything. Gravity separates gases slightly through the deep firn column, enriching heavier molecules downward; this is measured, not assumed, using the nitrogen isotope ratio in the trapped air, which should be constant in the atmosphere and so reveals exactly how much settling occurred. And below roughly a kilometre, pressure forces the bubbles to disappear entirely into the ice crystal lattice as clathrate hydrates — the gas is still there and still measurable, but you can no longer read it bubble by bubble.

So how is any of it dated? Near the surface, by counting annual layers, which show up in dust, chemistry, and isotopes with the seasons. Deeper, layers thin under their own weight and counting fails, so flow models take over, anchored by markers with independent dates — sulfate spikes from known volcanic eruptions, which also let cores from opposite ends of the Earth be aligned to each other.

And why believe the whole chain? Because of an overlap that could have failed and did not. At high-accumulation Antarctic sites such as Law Dome, the trapped air is young enough that the ice core record runs right up to the era of direct atmospheric monitoring, which has been continuous since 1958. The two independent methods measure the same decades, and they agree. That agreement is the reason to trust the parts of the record where no instrument exists — the method has been checked against instruments precisely where checking was possible.

The misconception worth correcting outright is that a bubble and the ice around it are contemporaries. They are not, and treating them as such would systematically misdate every gas measurement — by centuries at the sites that reach furthest back. A single core therefore carries two timescales that must be reconciled, and the uncertainty in that reconciliation is often larger than the uncertainty in either date on its own. It matters enormously for questions of sequence, such as whether a temperature change led or followed a carbon dioxide change across a glacial transition, since that question is precisely a comparison between the ice clock and the gas clock.

It is worth setting this beside tree rings, the other great annually-resolved archive, because the contrast is instructive rather than merely different. A tree ring holds its signal in the wood laid down that same year, so archive and signal share a date, and crossdating pins that date to the calendar exactly. An ice core splits them: the solid record and the gaseous record in the same piece of ice belong to different times. In exchange, the ice gets something no tree can offer — an actual physical sample of the atmosphere, and a record reaching back through eight glacial cycles at Dome C, where the oldest trees manage a few thousand years.

• How the water molecules of the ice itself, through their oxygen and hydrogen isotopes, record temperature independently of the gases inside it.

Every term the collection defines is gathered in the glossary.