During the warm periods between recent ice ages, temperatures in Antarctica reached substantially higher levels than scientists had previously thought. This conclusion, based on ice-core studies, implies that East Antarctica is more sensitive than it seemed to global warming. Previous estimates suggested that peak temperatures during the warmest interglacial periods — which occurred at around 125,000, 240,000 and 340,000 years ago — were about three degrees higher than they are today. But a team led by Louise Sime of the British Antarctic Survey in Cambridge, UK, concludes that Antarctica was actually around six degrees warmer.
The team based its analysis on ratios of oxygen and hydrogen isotopes in ice cores drilled in East Antarctica. Warmer air temperatures cause more water containing the heavier isotopes oxygen-18 or deuterium to evaporate from the surrounding ocean. Once it falls on inland Antarctica and packs into ice, it gives climate researchers a proxy of local temperatures at that time. Climate reconstructions usually assume a simple linear relationship between these isotope ratios and temperature. But Sime's team says that although that relationship holds up for the cold glacial periods, it does not work so well during the warmer interglacials.
The scientists measured the isotope ratios in three ice cores from across East Antarctica, each of which dates back to at least 340,000 years ago. They then compared those results with predicted isotope distributions derived from a global climate model.
They found that higher average temperatures were required to reconcile observations and model experiments. "The available evidence only fits together if we assume peak temperatures around six degrees above current values," says Sime. "We didn't expect this at all." The team believes that the relationship between temperature and the isotopic composition of water vapour changes as climate warms. For example, the isotopic signatures of ice cores depend on the seasonal distribution of precipitation. A change in the time of year when most snow falls could lead to biases in temperature reconstruction, says Sime, whose team reports its findings in Nature 1.
EDIT
http://www.nature.com/news/2009/091118/full/news.2009.1094.html