Oxygen isotope proxy errors corrected in Greenland ice cores

Posted on September 22, 2009


Scientists, mariners, and weather hobbyists started directly measuring temperature with thermometers globally in the late 1800s. When modern climatologists want temperature data farther back in time than those first global measurements, they have to use things called “proxies.” A proxy for temperature is something that, when calibrated properly, indirectly measures temperature. The most common proxies that are used as temperature stand-ins tend to be tree rings, the amount of an oxygen isotope in ice cores, and coral growth rings.

There are a couple of problems with proxies, however. The first problem is that scientists have to develop an appropriate and accurate calibration method to convert the width of a tree ring to an average annual or summer temperature. The second problem is that a given proxy may well be influenced by other factors beyond temperature, and so calibrating the proxy becomes a difficult and potentially error-prone process. For example, tree rings are a proxy for both temperature and moisture, and so any climatologist who wants to extract just the temperature information needs to discover a way to independently estimate the effect of moisture changes on the tree ring before the effect of temperature on the tree ring can be accurately determined.

A new study published September 17th as a letter in the journal Nature describes a new method to compensate for proxy changes due to elevation in the Greenland ice sheet (GIS) during the Holocene (the present geologic epoch, starting about 12,000 years ago).

This basic problem is that the oxygen isotope used as a temperature proxy in ice cores, 18O, varies with regional temperature, the body of water from which the snow originated, the path the water vapor traveled from its source to where it falls as snow, what season the snow was deposited in, how close the ice core is to the pole, and even the altitude at which the snow fell. A great deal of science has been done to understand how 18O changes with all of those factors, but sometimes errors creep in anyway. In this case, a previously un-corrected error in ice core 18O data from the GIS had confounded understanding the response of the GIS to warming during the Holocene.

The main problem is that the GIS used to be a lot thicker than it is today. At the start of the Holocene, the Earth was transitioning from an ice age to an interglacial, and as a result the Earth was quickly warming and sea levels were rising as a result of melting ice caps. And during that period, the GIS shrank and thinned, effectively lowering the elevation of the GIS at the same time. What this means is that there was some unknown amount of error in the 18O isotope signature in the ice cores, and that error was causing all sorts of problems. The image below, specifically part “b,” illustrates how the 18O varied significantly from one part of the GIS to another (part “a” shows where the cores were drilled, and part “c” will become important in a minute).


The authors set out to find sites that they could use to correct the 18O elevation effects in the GIS. They found two (Agassiz and Renland) that they could justify as being understood well enough to correct the other four sites in the GIS cores. And when they calibrated the 18O for those two cores for their known elevation and distance from the North Pole, they got the image above, part “c.” They’re not exactly the same, but they were close enough to use as calibration sources for the other four ice cores.

When the authors calibrated the other four ice core locations, they discovered that there had been significant elevation changes as the GIS thinned during the transition to the Holocene. In addition, the authors compared their new corrected proxy information to an elevation proxy, specifically the total gas content held in the ice. They compared the estimated elevation changes at two sites (GRIP and Camp Century) using the two different methods and discovered that they were qualitatively and quantitatively similar.

The authors also compared their data against models of the GIS and found that the ice sheet models did not accurately estimate the changes in elevation. In general, the models underestimated the change in GIS elevation, and thus the rate and amount of ice melt. So in addition to correcting a significant bias in the 18O temperature proxy record, the author also found that the GIS is more sensitive to temperature changes than expected.

It is therefore entirely possible that a future temperature increase of a few degrees Celsius in Greenland will result in GIS mass loss and contribution to sea level change [that is] larger than previously projected.

Thanks to Ubertramp who was kind enough to help me obtain a copy of the paper.

Image Credit

Posted in: Uncategorized