On the origin of negative deuterium excess observed in snow and ice samples from McMurdo Dry Valleys and Allan Hills Blue Ice Areas, East Antarctica

The oxygen and hydrogen isotopic composition (18O and D values) in snow and ice have long been utilized to reconstruct past temperatures of polar regions. Embedded in this approach is an assumption that post-depositional processes such as sublimation do not impact the isotopic composition of snowan assumption that has recently been questioned. In areas near the McMurdo Dry Valleys in Antarctica, for example, where the accumulation rate is smaller than 0.01 m yr-1, surface snow and ice samples are characterized by a negative deuterium excess (D - 8*18O). This unique phenomenon, only observed near the Dry Valleys, raises the possibility that sublimation may be capable of altering the isotopic composition of the surface snow. Here we use an isotope-enabled general circulation model and an ice physics model to investigate the origin of negative deuterium excess values observed in Dry-Valleys surface snow and ice samples. Our results suggest that negative deuterium excess values can only arise from precipitation if all the moisture is sourced from the Southern Ocean. However, the model results show that moisture sourced from oceans north of 55 S contributes significantly (~60%) to precipitation in Antarctica in the present day. While this contribution likely changes in different climates, it is likely a persistent feature of Antarctic precipitation. We thus propose that sublimation must have occurred to yield the negative deuterium excess values in ice at the Allan Hills, which is located outside the Dry Valleys. We calculate that under present-day conditions, 3 to 24% of the snow is lost due to sublimation. Because a higher fraction of snow is expected to be sublimated when accumulation rates are lower, the enrichment of 18O and D values due to sublimation will be higher during past cold periods than at present. This sensitivity of the snow isotopic composition to accumulation rates and sublimation implies that the cooling in glacial periods could be underestimated if the modern spatial relationship between temperature and precipitation D and 18O values is applied in cases where no independent temperature calibration is available.