Investigating the spatial representativeness of Antarctic ice cores: A comparison of ice core and radar-derived surface mass balance

Abstract. Surface mass balance (SMB) over the Antarctic Ice Sheet must be better understood to document current Antarctic contribution to sea-level rise. Field point data using snow stakes and ice cores are often used to evaluate the state of the ice sheet's mass balance as well as to validate SMB derived from regional climate models, which are then used to produce future climate projections. However, spatial representativeness of individual point data remains largely unknown, particularly in the coastal regions of Antarctica with highly variable terrains. Here, we compare ice core data collected at the summit of eight ice rises along the coast of Dronning Maud Land, as well as at the Dome Fuji site, and shallow ice-penetrating radar data over these regions. Shallow radar data has the advantage of being spatially extensive with a temporal resolution that varies between annual and sub-decadal resolution from which we can derive a SMB record over the entire radar survey. This comparison allows us therefore to evaluate the spatial variability of SMB and the spatial representativeness of ice-core derived SMB. We found that ice core mean SMB is very local and the difference with radar-derived SMB increases in a logarithmic-fashion as the surface covered by the radar data increases, with for most ice rises a plateau ~1–2 km away from the ice crest where there are strong wind-topography interactions, and ~10 km where the ice shelves begin. The relative uncertainty in measuring SMB also increases rapidly as we move away from the ice core sites. Five of our ice rise sites show a strong spatial representativeness in terms of temporal variability, while the other three sites show it is limited to a surface areas between 20–120 km². The Dome Fuji site on the other hand shows a small difference between pointwise and area mean SMB, as well as a strong spatial representativeness in terms of temporal variability. We found no simple parameterization that could represent the spatial variability observed at all the sites. However, these data clearly indicate that local spatial SMB variability must be considered when assessing mass balance as well as comparing modeled SMB values to point field data.