Historical Geology/Ice cores
In this article we shall look at how core samples from ice can be used to give us information about paleoclimates. The reader may find it useful to look back at the main article on glaciers before reading further.
At any location where snow falls but does not melt (or at least does not completely melt) before the next year's snowfall, the snow will accumulate, and as each year's snow is buried by the further snowfall of succeeding years it compacts from loose snow to permeable firn to impermeable ice, at which point it is said to undergo closure. You should remember from the article on glaciers that any point at which snow accumulates like this must inevitably become the accumulation zone of a glacier.
Ice formed from summer snow is lighter and less dense than ice formed from winter snow; as a result, if the rate of accumulation is more than about 4 cm/year we get what in effect are varves formed from snow. At least in the upper part of the glacier, these are plainly visible if we take a core sample, as you can see in the photograph below right.
Ice core data
The fact that the ice cores contain visible annual layers means that just as with varves in lakes, we can count them down from the top, and assign a year to each one.
Looking at the thickness of each layer, we can quantify the amount of snowfall, or at least unmelted snowfall, in the summer and winter of each year. What is perhaps more interesting, we can measure the 16O/18O ratio of the water, which acts as a climatic indicator for reasons discussed in the article on scleroclimatology; we can also measure the 1H/2H ratio of the water, which acts as a climatic indicator for the same reason: water molecules containing the 2H isotope are slightly heavier, and so evaporate less readily.
Besides these data, when the permeable firn turns to impermeable ice, bubbles of the atmosphere become trapped within the ice. This allows us to analyze the past composition of the atmosphere, and quantify gasses which affect the climate, such as carbon dioxide (CO2) methane (CH4) and sulfur dioxide (SO2).
Also, as with ordinary sediment, the ice cores will contain windborne particles including volcanic ash, pollen, and loess (dust produced by the action of glaciers).
How do we know?
We can see the annual formation of layers in ice, and we also understand the mechanism behind this in terms of seasonal variation. For the layers not to form at all in a given year, there would have to be a complete cessation of snowfall, which is very unlikely; for more than one layer to form per year we would need more than one warm period per year, which is almost impossible.
We can also verify that the data preserved in the layers reflects the climate by comparing the ice core record with direct measurements of climate made over the last few centuries, and with other climatic proxies.
To this we can add that the methods outlined in this article should work; they are based, after all, on very simple physical principles. Is it even conceivable that in times gone by water composed of heavy isotopes evaporated more easily than water composed of light isotopes? Or that once upon a time when firn underwent closure it trapped bubbles of something other than the atmosphere?
So the analysis of ice core data ought to work in principle, and comparison of the results with other data confirms that it does work in practice.
Difficulties with the method
Because firn doesn't turn to ice immediately, and because the air bubbles aren't trapped until it does, the year we calculate for the age of a layer will be different from the age of the atmosphere trapped in that layer. The difference (closure time) between the age of the gas and the age of the ice can be as much as 7,000 years, as is the case in ice cores from Vostok, or as little as 30 years at Law Dome.
Mathematical methods can be used to recover data on a finer scale than the closure time; however, such methods rely on knowing the closure time. Now, the climate has changed in the past (otherwise paleoclimatology would be a completely unnecessary science) and so it seems likely that closure time will have varied from time to time in the past just as it varies from place to place in the present. This introduces an element of uncertainty into the data.
Another problem is that at greater depths the annual layers become more and more indistinct, until in some cases they can't be made out at all. One can try in such cases to estimate the age of the buried ice by measuring its depth and the accumulation at the location in question, but again this would only work perfectly if the climate, or at least the snowfall, had remained constant each year. Where the ice cores contain volcanic ash, this can be dated by radiometric methods, allowing us to put correct dates on the layers in which they are found, but only up to the accuracy of the radiometric methods used.
There is one final weakness inherent in the method. By definition, any point where the annual rate of snowfall exceeds the rate it melts must be the accumulation zone of a glacier; in the case of the Greenland and Antarctic ice sheets that most interest paleoclimatologists, the accumulation zones of continental glaciers. And a glacier must flow out from its zone or zones of accumulation to a zone or zones where it ablates. In short, the record in the ice is progressively being destroyed as well as created. This happens relatively fast by geological standards, providing us with a record that can be measured in hundreds of thousands of years; compare this with proxies based on marine sediment, which is only destroyed by the much slower process of subduction.
However, ice cores remain valuable because the ice does trap air in its pore spaces, providing us with a continuous record of the composition of the atmosphere.