Decoupling of δ18O from surface temperature in Antarctica in an ensemble of historical simulations

Stable water isotopes recorded in Antarctic ice cores have traditionally been used to infer past surface air temperatures (SATs). During the historical period (1850 onward), observational data and good-quality ice core records overlap, yielding an opportunity to investigate key relationships between ice core stable water isotope (δ18O) measurements and the Antarctic climate. We present a new ensemble of climate model simulations covering 1851–2004 using the UK Met Office HadCM3 general circulation model equipped with stable water isotopes. Our ensemble captures observed historical SAT and precipitation trends and weak δ18O trends. The weak δ18O trends mean there is no significant relationship between SAT and δ18O over one-third of Antarctica, and also half of our considered ice core sites, though relationships are stronger when using regional averages. The strongest regional relationships occur in the West Antarctic Ice Sheet (WAIS) region. This decoupling between SAT and δ18O occurs primarily because of the impact of autumnal sea ice loss during the simulated warming. The warming and sea ice loss are associated with (i) changes in near-coastal air mass intrusions (synoptic effects) induced by changes in the large-scale circulation and/or sea ice; (ii) direct sea-ice-driven changes in moisture pathways (especially lengths) to Antarctica; and (iii) precipitation seasonality changes, again mostly driven by sea ice changes. Consequently, when reconstructing temperatures over these timescales, changes in sea ice need to be considered, both to determine the most appropriate SAT and δ18O relationship and to understand how uncertainties affect the inference of past temperature from ice core δ18O measurements.