Estimating small-scale snow depth and ice thickness from total freeboard for East Antarctic sea ice

Deriving the snow depth on Antarctic sea ice is a key factor in estimating sea-ice thickness distributions from space or airborne altimeters. Using a linear regression to model snow depth from observed ‘total freeboard’, or the snow/ice surface elevation relative to sea level is an efficient and promising method for the estimation of snow depth for instruments which only detect the uppermost surface of the sea-ice conglomerate (e.g. laser altimetry). However the Antarctic pack-ice zone is subject to substantial variability due to synoptic-scale weather forcing. Ice formation, motion and melt undergo large spatio-temporal variability throughout the year. In this paper we estimate snow depth from total freeboard for the ARISE (2003), SIPEX (2007) and SIPEX-II (2012) research voyages to the East Antarctic pack-ice zone. Using in situ data we investigate variability in snow depth and show that for East Antarctica, relationships between snow depth and total freeboard vary between each voyage. At a resolution of metres to tens of metres, we show how regression-based snow-depth models track total freeboard and generally over-estimate snow depth, especially on highly deformed sea ice and at sites where ice freeboard makes a substantial contribution to total freeboard. For a set of 3192 records we obtain an in situ mean snow depth of 0.21m (σ=0.19m). Using a regression model derived from all in situ points we obtain the same mean, with a slightly lower variability (σ=0.16m). Using voyage-specific subsets of the data to derive regression models and estimate snow depth, mean snow depths ranged from 0.19m (model derived from SIPEX observations) to 0.25m (model derived from SIPEX-II observations). While small, these discrepancies impact ice thickness estimation using the assumption of hydrostatic equilibrium. Mean in situ ice thickness for all samples is 1.44m (σ=1.19m). Using empirical models for snow depth, ice thickness varies from 1.0 to 1.8m with the best match to the in situ mean given when snow depth is derived using a snow depth model from all observations (1.53m, σ=1.55m). However, mean values only tell part of the story when investigating the sea-ice thickness distribution. Here we explicitly show how modelling snow depth and ice thickness based on a total freeboard signal compares with in situ observations. This provides insight into the confidence we place in ice thickness distributions derived using a total freeboard signal and empirically-derived models f