Sea Ice Roughness Overlooked as a Key Source of Uncertainty in CryoSat‐2 Ice Freeboard Retrievals

Abstract

ESA's CryoSat-2 has transformed the way we monitor Arctic sea ice, providing routine measurements of the ice thickness with near basin-wide coverage. Past studies have shown that uncertainties in the sea ice thickness retrievals can be introduced at several steps of the processing chain, for instance, in the estimation of snow depth, and snow and sea ice densities. Here, we apply a new physical model to CryoSat-2, which further reveals sea ice surface roughness as a key overlooked feature of the conventional retrieval process. High-resolution airborne observations demonstrate that snow and sea ice surface topography can be better characterized by a lognormal distribution, which varies based on the ice age and surface roughness within a CryoSat-2 footprint, than a Gaussian distribution. Based on these observations, we perform a set of simulations for the CryoSat-2 echo waveform over "virtual" sea ice surfaces with a range of roughness and radar backscattering configurations. By accounting for the variable roughness, our new lognormal retracker produces sea ice freeboards that compare well with those derived from NASA's Operation IceBridge airborne data and extends the capability of CryoSat-2 to profile the thinnest/smoothest sea ice and thickest/roughest ice. Our results indicate that the variable ice surface roughness contributes a systematic uncertainty in sea ice thickness of up to 20% over first-year ice and 30% over multiyear ice, representing one of the principal sources of pan-Arctic sea ice thickness uncertainty. Plain Language Summary We have developed a new way of measuring sea ice thickness in the Arctic, by comparing real and simulated data from the European Space Agency satellite: CryoSat-2. Our simulations are guided by aircraft observations that demonstrate sea ice has distinct patterns of surface roughness. Traditional methods ignore or misrepresent the surface roughness, which reduces our confidence in the measured ice thickness by around one third. If we account for the roughness, however, we can extend the capability of Cryosat-2 for measuring both the thinnest smoothest sea ice and thickest roughest ice. These improvements will boost our confidence in derived estimates of the Arctic Ocean's sea ice volume budget and freshwater fluxes, while enhancing the accuracy of sea ice forecasts primed with satellite data.

The 2017 Reversal of the Beaufort Gyre: Can Dynamic Thickening of a Seasonal Ice Cover During a Reversal Limit Summer Ice Melt in the Beaufort Sea?

“…The sea ice freeboard is estimated from the elevation difference between ice floes and sea surface height interpolated between leads. Sea ice freeboard distributions obtained with LARM compare closely to those obtained from coincident airborne data from NASAs Operation IceBridge (Landy et al., 2020).…”

“…Sea ice freeboard and surface roughness ( σ ) observations were obtained from the ESA CryoSat‐2 satellite, using the Lognormal Altimeter Retracker Model (LARM) algorithm described by Landy et al. (2020). The LARM algorithm is based on simulations of the CryoSat‐2 waveform performed with a physical model for the SAR altimeter echo backscattered from sea ice (Landy et al., 2019).…”

Multi-peak Retracking of CryoSat-2 SARIn Waveforms Over Arctic Sea Ice

“…However, these results show that, when using a 50% threshold retracker, the across-track location corresponding to the waveform retracking point slightly deviates, on average, from the satellite nadir. This confirms that threshold-based empirical retrackers might not always pick a point on the waveform corresponding to the exact satellite nadir, as the location of this point is influenced by surface roughness at the footprint scale [33], [51], [52]. Point estimates of sea ice freeboard obtained from ONC-corrected elevations are, therefore, likely to be always more accurate than uncorrected estimates, as the ONC takes care of biases introduced by empirical retrackers [15].…”

Ensemble-based estimation of sea-ice volume variations in the Baffin Bay

“…Because in situ observations of SIT are very scarce in Baffin Bay, a locally merged satellite SIT (Sat-merged SIT) data set is utilized to calculate the SIV variations during the freezing season, since this data set was already used to estimate the sea-ice variations in the eastern Canadian Arctic including Baffin Bay. These Sat-merged SIT data are calculated from CryoSat-2 radar freeboards (accessed from the European Space Agency) and PMW snow depth (available from NSIDC at https://nsidc.org/data/NSIDC-0032/versions/ 2, last access: 28 October 2020) and then merged with SMOS SIT (available from the University of Hamburg at https: //icdc.cen.uni-hamburg.de/en/l3c-smos-sit.html, last access: 28 October 2020), where the mean CryoSat-2 thickness is < 1 m. More details about this data set can be found in Landy et al (2017Landy et al ( , 2019Landy et al ( , 2020.…”

“…For instance, SMOS SIT is underestimated because (1) SMOS only provides valid SIT for ice thinner than 1 m, and (2) the 100 % ice concentration assumption during the data retrieval is not fulfilled (Tian-Kunze et al, 2014;Tietsche et al, 2018). To address the challenging estimation of sea-ice volume variations in Baffin Bay, locally merged satellite SIT data (Landy et al, 2017(Landy et al, , 2019(Landy et al, , 2020 and three sea ice-ocean models driven by atmospheric reanalysis are employed, namely the sufficiently well validated combined model and satellite sea-ice thickness (CMST), the widely used Pan-Arctic Ice Ocean Modeling and Assimilation System (PIOMAS), and a version of the North Atlantic/Arctic Ocean Sea Ice Model (NAOSIM) with optimized parameters. Because very few in situ observations can be used for validation in Baffin Bay, we carry out an inter-comparison between CMST, NAOSIM, PIOMAS, the Towards an Operational Prediction system of the North Atlantic and European coastal Zones (TOPAZ4) and the merged satellite SIT of Landy et al (2017) (named Sat-merged SIT hereafter).…”