The Cryosphere (Mar 2022)

Overestimation and adjustment of Antarctic ice flow velocity fields reconstructed from historical satellite imagery

  • R. Li,
  • R. Li,
  • Y. Cheng,
  • Y. Cheng,
  • H. Cui,
  • H. Cui,
  • M. Xia,
  • M. Xia,
  • X. Yuan,
  • X. Yuan,
  • Z. Li,
  • Z. Li,
  • S. Luo,
  • S. Luo,
  • G. Qiao,
  • G. Qiao

DOI
https://doi.org/10.5194/tc-16-737-2022
Journal volume & issue
Vol. 16
pp. 737 – 760

Abstract

Read online

Antarctic ice velocity maps describe the ice flow dynamics of the ice sheet and are one of the primary components used to estimate the Antarctic mass balance and contribution to global sea level changes. In comparison to velocity maps derived from recent satellite images of monthly to weekly time spans, historical maps, from before the 1990s, generally cover longer time spans, e.g., over 10 years, due to the scarce spatial and temporal coverage of earlier satellite image data. We found velocity overestimations (OEs) in such long-span maps that can be mainly attributed to velocity gradients and time span of the images used. In general, they are less significant in slow-flowing grounded regions with low spatial accelerations. Instead, they take effect in places of high ice dynamics, for example, near grounding lines and often in ice shelf fronts. Velocities in these areas are important for estimating ice sheet mass balance and analyzing ice shelf instability. We propose an innovative Lagrangian velocity-based method for OE correction without the use of field observations or additional image data. The method is validated by using a set of ground truth velocity maps for the Totten Glacier and Pine Island Glacier which are produced from high-quality Landsat 8 images from 2013 to 2020. Subsequently, the validated method is applied to a historical velocity map of the David Glacier region from images from 1972–1989 acquired during Landsat 1, 4, and 5 satellite missions. It is demonstrated that velocity overestimations of up to 39 m a−1 for David Glacier and 195 m a−1 for Pine Island Glacier can be effectively corrected. Furthermore, temporal acceleration information, e.g., on basal melting and calving activities, is preserved in the corrected velocity maps and can be used for long-term ice flow dynamics analysis. Our experiment results in the Pine Island Glacier (PIG) show that OEs of a 15-year span can reach up to 1300 m a−1 along the grounding line and cause an overestimated grounding line (GL) flux of 11.5 Gt a−1 if not corrected. The magnitudes of the OEs contained in both velocity and mass balance estimates are significant. When used alongside recent velocity maps of 1990s–2010s, they may lead to underestimated long-term changes for assessment and forecast modeling of the global climate change impact on the Antarctic ice sheet. Therefore, the OEs in the long-span historical maps must be seriously examined and corrected. We recommend that overestimations of more than the velocity mapping uncertainty (1σ) be corrected. This velocity overestimation correction method can be applied to the production of regional and ice-sheet-wide historical velocity maps from long-term satellite images.