Enhancing Arctic sea surface height and sea ice freeboard mapping with off-track leads

Abstract:

With the launch of a pair of satellite altimeters focused on the polar regions – ESA’s CryoSat-2 in 2010 and NASA’s ICESat-2 in 2018 – we are now able to routinely monitor the sea ice thickness across almost the entire Arctic Ocean. One of the key uncertainties remaining in the sea ice freeboard processing chain is the error on the sea surface height (SSH) estimated at ice floes. For CryoSat-2, the SSH is conventionally interpolated from lead samples located at the closest points along the altimeter’s orbital track. However, along-track distances to the nearest lead can often be well in excess of 100 km where ice concentrations are high, producing uncertainties on the derived sea ice freeboard that can reach more than 50%. Furthermore, SSH estimates from leads on a single track are generally not independent so errors do not reduce with accumulation.

Here we use an objective mapping approach to determine, for every sea ice floe along the CryoSat-2 track, the optimal SSH from all proximal lead samples located both on and off the orbital track. The patterns of the Arctic sea level anomaly’s (SLA) spatial and temporal decorrelation length scales are obtained by analysing the covariance of the SLA signal from all CryoSat-2 lead observations between 2010 and 2019. We compare these decorrelation scales with the expected Rossby radius of vertical SSH deformation and scales obtained from eddy-permitting simulations with the coupled NEMO-LIM2 ocean model. The derived length scales are combined with random and systematic errors on the SSH signal and the Arctic Ocean mean geostrophic circulation field to constrain the optimal SSH and best error estimate at each ice floe.

By ‘using’ all available lead samples (tens-hundreds of times more than on the orbital track alone), we can characterise the SLA at much shorter wavelengths than the conventional approach. The rms of the SLA measured at orbital crossovers, the estimated sea level uncertainty, and the accuracy of sea ice freeboards derived from the reprocessed SLA all show significant improvements. Additionally, by exploiting a vastly greater number of lead samples, the influence of a single lead elevation (and thus an off-nadir error, for instance) on the derived sea ice freeboard is considerably reduced.