Sea ice thickness estimations from ICESat Altimetry over the Bellingshausen and Amundsen Seas, 2003–2009

Abstract

Sea ice thicknesses derived from NASA's Ice, Cloud, and Land Elevation Satellite (ICESat) altimetry data are examined using two different approaches, buoyancy and empirical equations, and at two spatial scales—ICESat footprint size (70 m diameter spot) and Advanced Microwave Scanning Radiometer (AMSR‐E) pixel size (12.5 km by 12.5 km) for the Bellingshausen and Amundsen Seas of west Antarctica. Ice thickness from the empirical equation shows reasonable spatial and temporal distribution of ice thickness from 2003 to 2009. Ice thickness from the buoyancy equation, however, additionally needing snow depth information derived from the AMSR‐E, shows an overestimation in terms of maximum, mean (+63% to 75%), and standard deviation while underestimation in modal thickness (−20%) as compared with those from the empirical equation approach. When ICESat snow freeboard is used as the snow depth in the buoyancy equation, i.e., the zero ice freeboard assumption, the derived ice thicknesses match well with those from the empirical equation approach, within 5% overall. The AMSR‐E, therefore, may underestimate snow depth and accounts for ~95% of the ice thickness overestimation as compared with the buoyancy approach. The empirical equation derived ice thickness shows a consistent asymmetrical distribution with a long tail to high values, and seasonal median values ranging from 0.8 to 1.4 m over the 2003–2009 period that are always larger than the corresponding modal values (0.6–1.1 m) and lower than the mean values (1.0–1.6 m), with standard deviation of 0.6–1.0 m. An overall increasing trend of 0.03 m/year of mean ice thickness is found from 2003 to 2009, although statistically insignificant (p = 0.11) at the 95% confidence level. Starting from autumn, a general picture of seasonal mean, modal, and median ice thickness increases progressively from autumn to spring and decreases from spring to the following autumn, when new thin ice dominates the ice thickness distribution. The asymmetric shape of the thickness distribution reflects the key role of ice deformation processes in the evolution of the thickness distribution. The statistical properties of the thickness distribution interannually (high range of mean thickness and standard deviation) indicate the variability of deformation processes. However, spring ice volume, the product of ice mean thickness and areal extent computed for the spring maximum, shows variability year to year but is primarily dominated by ice extent variability, with no increasing or decreasing trend over this record length. The dependence of the volume on the ice extent primarily suggests that ice thickness changes have also not covaried with the ice extent losses seen over the satellite record in this region, unlike the Arctic. These properties reflect the interactive processes of ice advection, thermodynamic growth and ice deformation that all substantially influence ice mass balance in the Bellingshausen‐Amundsen Seas region.