UC San Diego

Polar temperature is often inferred from water isotopes in ice cores. However, non-temperature effects on δ18O are important during the abrupt events of the last glacial period, such as changes in the seasonality of precipitation, the northward movement of the storm track, and the increase in accumulation. These effects complicate the interpretation of δ18O as a temperature proxy.Here, we present an independent surface temperature reconstruction, which allows us to test the relationship between δ18Oice and temperature, during Dansgaard-Oeschger event 8, 38.2 thousand yrs ago using new δ15N and δ40Ar data from the GISP2 ice core in Greenland. This temperature reconstruction relies on a new inversion of inert gas isotope data using generalized least-squares, and includes a robust uncertainty estimation.We find that both temperature and δ18O increased in two steps of 20 and 140 yrs, with an overall amplitude of 11.80 ± 1.8°C between the stadial and interstadial centennial-mean temperature. The coefficient α=dδO18/dT changes with each time-segment, which shows that non-temperature sources of fractionation have a significant contribution to the δ18O signal. When measured on century-averaged values, we find that α=dδO18/dT=0.32±0.06‰/°C, which is similar to the glacial/Holocene value of 0.328‰/°C. © 2014 Elsevier B.V.