How Ice Cores Record Past Climates

When snow falls on the high interior of Antarctica or Greenland, it almost never melts. Each year a fresh layer settles onto the last, and the weight of accumulating snow above slowly compresses what lies beneath. Loose flakes pack into granular firn, then into solid ice, and as they do, the air between the grains is sealed off into tiny bubbles. A column of this ice, drilled out as a cylinder several kilometers long, is an archive: the deepest ice in the EPICA Dome C core in East Antarctica is roughly 800,000 years old, and the bubbles inside it hold samples of the atmosphere that existed when woolly mammoths still walked Europe.

The most direct signal is the air itself. Once a bubble closes off, the gas inside is effectively trapped. By crushing slices of ancient ice in a vacuum and feeding the released gas into a mass spectrometer, researchers measure the concentrations of carbon dioxide, methane, and nitrous oxide breath by breath through deep time. The results are unambiguous on one point: across the last 800,000 years, atmospheric CO2 oscillated between roughly 180 parts per million during ice ages and 280 during warm interglacials, and never approached today's values until the industrial era.

The ice itself records temperature, but indirectly, through the chemistry of the water molecules that compose it. Water made with the heavier isotope oxygen-18 evaporates slightly less readily and condenses slightly more readily than ordinary water made with oxygen-16. When the climate is cold, the heavier molecules tend to rain out of air masses long before they reach the polar interior, leaving the snow that does fall there isotopically light. The ratio of oxygen-18 to oxygen-16 in a layer of ice, written as delta-O-18, therefore tracks the temperature at which that snow originally fell. A similar logic applies to deuterium, the heavy isotope of hydrogen, which gives a second independent temperature estimate from the same ice.

A core preserves more than gas and water. Wind-borne dust settles onto the ice each year and is buried with it, and the amount of dust in a layer reveals how arid and windy the source continents were at the time; glacial periods are consistently dustier than warm ones. Sulfate spikes pinpoint major volcanic eruptions, which serve as time markers visible across multiple cores. In layers from the last few thousand years, traces of lead from Roman smelting and, later, from leaded gasoline appear as a faint chemical fingerprint of human industry.

Reading these signals requires a chronology, and dating the ice is its own discipline. In the upper portions of a core, annual layers can often be counted directly, like tree rings, because summer and winter snow differ in dust content and isotopic composition. Deeper down, the layers thin under their own weight until they blur together, and dating relies instead on flow models, on matching volcanic horizons between cores, and on tying the trapped methane record to other dated archives such as cave deposits.

Two features of this archive deserve emphasis. First, the gas trapped in a bubble is younger than the ice surrounding it, sometimes by centuries, because the bubble does not seal until the firn fully closes hundreds of meters below the surface. Any comparison between a temperature signal in the ice and a gas signal in the bubbles must correct for this offset. Second, ice cores from Greenland and Antarctica do not always tell the same story: abrupt warmings recorded in Greenland during the last ice age appear in Antarctica as slower, opposite-signed changes, a hemispheric seesaw that has reshaped how scientists think about ocean circulation.

The power of the method lies in this layering of evidence. A single core yields temperature, atmospheric composition, volcanic history, and continental dustiness from the same depth, the same year. When several cores agree, the past stops being a guess.