The clock in the ice

Snow that falls on the high plateau of Antarctica does not melt. The following winter buries it, and the winter after that buries the layer above. Under its own weight the snow is pressed into ice, and the air between the original snowflakes is sealed into bubbles. What remains is a column in which every year sits on top of the year before, and in which a sample of the atmosphere of that year is still held. Drilling down through such a column is the closest thing earth science has to reading a diary kept by somebody else.

Near the surface the years can simply be counted. Snow that falls in summer has larger crystals than snow that falls in winter, and it carries more dust, so the column is banded much as a tree trunk is banded. Deeper down the bands are squeezed too thin to separate by eye, and other markers take over. A large volcanic eruption lays a thin film of sulphate across the whole continent within a year or two, and because the dates of many eruptions are known from written and geological records elsewhere, each film becomes a fixed point against which the ice above and below can be placed.

The bubbles are the prize. Most of what is known about past climate is inferred from something else that happened to be living at the time: a shell, a pollen grain, a ring of wood. Ice is different, because the gas held in the bubbles is the old air itself, and the proportion of carbon dioxide and methane in it can be measured directly rather than estimated. Temperature has to be worked out less directly, from the ratio of two forms of oxygen in the frozen water, which varies with how cold the air was at the moment the snow formed.

Getting the ice out is slow work. A drilling station has to be built where the ice is thick and has not flowed sideways, which in practice means the middle of the plateau, a thousand kilometres from the nearest coast. The borehole is filled with a dense fluid so that the pressure of the surrounding ice does not squeeze it shut between seasons. Cores come to the surface in sections roughly a metre long, are logged and cut, and are shipped out still frozen, so that a single deep hole can occupy one team for several summers.

There is a limit to how far back the method reaches. Ice is not rigid. Under pressure it spreads outwards, and near the bottom of the sheet the oldest layers are stretched, folded and sometimes turned over, so that age no longer increases neatly with depth. Worse, heat rising from the rock beneath can melt the lowest ice away altogether, which removes the record rather than merely disturbing it.

Even so, the ambition is to go older. Cores already drilled reach back something like eight hundred thousand years, which covers the last eight swings between glacial and warm periods, and several teams are now surveying for a site where ice one and a half million years old may have survived in place. The attraction of that particular age is a change of rhythm: ice ages before it arrived roughly every forty thousand years, and after it roughly every hundred thousand, and nobody yet agrees why.