Sea ice thickness is an important variable for sea ice monitoring and climate projections. Although observations of ice extent and concentration have been available since late 1978, the same is not true for ice thickness. Because of the high interannual variability of the ice pack, climate trends in thickness and volume can only be observed over long time series. The earliest measurements of sea ice thickness (as multi-year averages for ERS-1 and ERS-2) were published in 2003 (Laxon et al 2003) (but these results have never been replicated). Today, the longest series go back to the winter of 2002/2003 (the beginning of the Envisat mission).

The main difficulty in going back in time is linked to the difference in altimeter generation and processing. The launch of the first SAR altimeter aboard CryoSat-2 made it possible to obtain the first consistent measurements of freeboards measurements, thanks to its small footprint (about 5km2). Prior to CryoSat-2, space altimeters were in low-resolution mode (LRM), with a footprint of the order of 150km2 overlapping heterogeneous surface types. To improve Envisat measurements, different calibration methodologies relative to CryoSat-2 have been proposed taking advantage of the common flight period (2010-2012) (Guerreiro et al 2017, Paul et al 2018, Tilling et al 2019, etc.).

We propose in this study a more general and robust method based on a neural network taking into account the state of the ice, from its surface roughness to its type. In order to extend the coverage to the beginning of the polar altimeter era (1993), we extended this method to ERS measurements by taking advantage of the common period with Envisat. Therefore, the measurements are successively corrected with reference to the most accurate CryoSat-2 measurements, offering a homogeneous series over nearly 30 years.

If ERS thickness measurements have not been reproduced, it is also because of the pulse-blurring effect due to instabilities of the tracker board, which must be corrected. The methodology used in the study to overcome this, is the interpretation of the N.Peacock (2004) approach.

We are finally able to provide a radar freeboard product for the Arctic between 50°N and 82.5°N over almost 30 winters (up to 88°N for Cryosat-2). The freeboard product is given with corresponding uncertainties using a Monte Carlo methodology to propagate all mission uncertainties through the neural network. The time series is finally compared to numerous in situ datasets, airborne measurements, or other products from other types of altimeters such as the ICE-Sat missions. Unless it is complicated to give precise conclusions due to the lack of long and homogeneous time series of snow depth, the comparisons show good consistency between field data and altimetry data.

This study was carried out within the framework of a CNES/CLS PhD fellowship and is partly included in the ESA FDR4ALT project.

Since the launch of CryoSat-2, SAR altimetry has demonstrated its ability to improve the exploitation
of radar altimetry over sea ice, thanks to the better along-track resolution, the non-overlapping
footprint and the multi-looking processing. The Fully-Focussed SAR (FF-SAR) [2] by its along-track
resolution reaching the theoretical limit (around 50cm) and its noise reduction multi-looking processing,
offers a great potential for capturing very narrow leads at sea-ice surfaces. If a lot of analysis
has been performed to demonstrate the value added of unfocused SAR technique over sea ice, very
few analyses have been led with the FF-SAR technique on this surface [1].

To this end, to take advantage on the Sentinel constellations of satellites, we have developed tools
and processings allowing to collocate Sentinel-1 SAR images and Sentinel-2 multispectral images with
Sentinel-3 and Sentinel-6 altimetry measurements. As published in [3], the results show a good agreement
on lead detection between the Long´ep´e’s algorithm on Sentinel-1 and the waveform classification
on the Sentinel-3A unfocused SAR waveforms. To evaluate the capability of the FF-SAR to detect
narrow leads, the idea is to compare FF-SAR with lead detection algorithm co-dated with Sentinel-2
images in order to evaluate the minimum size of lead detected by the algorithm, thanks to the better
pixel resolution of the MSI instrument (up to 10x10m).

In this study we will define a criterium of lead presence based on Sentinel-6 FF-SAR magnitude
squared coherence. A previous statistical analysis on Sentinel-3 FF-SAR data collocated with the
Sentinel-1 and Sentinel-2 lead detection has already consolidated this criterium, even if some limitations
due to lead replicas was sometimes giving Matthews correlation coefficient lower than with
UF-SAR. Replica should not be a problem anymore with Sentinel-6 MF interleaved mode of Poseidon-
4 altimeter (except the drift motion of the sea ices due to the time lag between observations that may
alter the results), first FF-SAR radargrams on sea-ice have already shown its great improvement for
visualisation of small leads structures.

Keywords— FFSAR, sea-ice, lead detection

References

[1] L. N. Connor, A. Egido, and T. W. K. Armitage. An Assessment of Sentinel-3A Measurements and Fully-
Focused SAR Processing over Arctic Sea Ice. 2018:C21A–08, Dec. 2018. Conference Name: AGU Fall Meeting
Abstracts ADS Bibcode: 2018AGUFM.C21A..08C.
[2] A. Egido and W. H. F. Smith. Fully Focused SAR Altimetry: Theory and Applications. IEEE Transactions
on Geoscience and Remote Sensing, 55(1):392–406, Jan. 2017.
[3] N. Long´ep´e, P. Thibaut, R. Vadaine, J.-C. Poisson, A. Guillot, F. Boy, N. Picot, and F. Borde. Comparative
Evaluation of Sea Ice Lead Detection Based on SAR Imagery and Altimeter Data. IEEE Transactions on
Geoscience and Remote Sensing, 57(6):4050–4061, June 2019. Conference Name: IEEE Transactions on
Geoscience and Remote Sensing.

National Aeronautics and Space Administration's (NASA's) Ice, Cloud, and land Elevation Satellite 2 (ICESat-2) mission has provided routine, very high-resolution estimates of surface height/type (the ATL07 product) and freeboard (the ATL10 product) across the Arctic and Southern Oceans since data collection started in October 2018.

In recent work we have estimated sea ice thickness across the entire Arctic Ocean from ATL10 freeboards, using snow depth and density estimates from the NASA Eulerian Snow on Sea Ice Model (NESOSIM) (Petty et al., 2020). These thickness estimates are now hosted at the National Snow and Ice Data Center (NSIDC, https://nsidc.org/data/IS2SITMOGR4). Additionally, we have produced new estimates of higher-level sea ice state estimates related to concentration and floe size (Petty et al., 2021).

Here we provide an overview of new release updates made to the underlying ATL07/ATL10 products (now at Release 005) that impact basin-scale estimates of freeboard, lead fraction and chord length, together with updates to the NESOSIM model (now at v1.1), and the subsequent impacts on our estimates of sea ice thickness, including updated comparisons to data collected by the original ICESat mission, Operation IceBridge and ESA’s CryoSat-2. Misclassified leads were removed from the freeboard algorithm in the third release (Release 003) of ICESat-2 freeboard data which significantly increased freeboards in January and April 2019 (and increased the fraction of low freeboards in November 2018). These changes significantly improved comparisons of sea ice thickness (lower mean biases and standard deviations, higher correlations) with thickness estimates produced from ESA’s CryoSat-2 (using the same input snow and ice density assumptions). NESOSIM v1.1 generally produces thicker snow than v1.0, although these changes result in a less significant impact on thickness compared to the Release 003 freeboard changes.

With now three (going on four) winters of data collected by ICESat-2 over the entire Arctic, we highlight interannual differences in the seasonal evolution of freeboard, thickness and floe size. We also explore possible causes of differences based on an analysis of near-surface atmospheric conditions (ERA5), ice drift (NSIDC) ice type (OSI SAF) and freeze-up estimates (passive microwave).

Finally, we explore the production of a joint thickness-floe size distribution from these data and discuss optimal model assimilation strategies as we seek to integrate ICESat-2 sea ice data into state-of-the-art sea ice climate model components.

References
Petty, A. A., N. T. Kurtz, R. Kwok, T. Markus, T. A. Neumann (2020), Winter Arctic sea ice thickness from ICESat‐2 freeboards, Journal of Geophysical Research: Oceans, 125, e2019JC015764. doi:10.1029/2019JC015764

Petty, A. A., M. Bagnardi, N. T. Kurtz, R. Tilling, S. Fons, T. Armitage, C. Horvat, R. Kwok (2021), Assessment of ICESat-2 sea ice surface classification with Sentinel-2 imagery: implications for freeboard and new estimates of lead and floe geometry Earth and Space Science, 8, e2020EA001491. doi:10.1029/2020EA001491.

Satellite altimetry provides the means to measure sea ice freeboard, an essential parameter to estimate ice thickness. Synthetic Aperture Radar (SAR) altimeters, such as the European Space Agency’s CryoSat-2 (2010-present) and Sentinel-3 (2016-present), have demonstrated to be tremendously valuable for that task, particularly, given their improved resolution along the satellite track. However, observations of sea ice are still challenging; in summer due to the presence of melt ponds, and in regions with numerous leads, which contaminate waveforms and complicate surface discrimination.

Fully focused SAR is a novel data processing technique that can further improve the resolution of observations over sea ice by using phase information from the standard SAR data to focus echoes along the satellite track. This allows a better representation of radar backscatter changes on the surface and therefore a better discrimination between leads and floes, which could in turn result in better freeboard and thickness estimates compared with standard data processing techniques.

Here we present an evaluation of high resolution CryoSat-2 data over Arctic sea ice with coincident NASA’s Operation IceBridge (OIB) data. During its spring Arctic campaigns of 2016, 2017, and 2018, the OIB mission, utilizing a multi-instrumented aircraft, collected spatially and temporally coincident data with CryoSat-2 in the eastern Beaufort Sea, a study area comprising large sea ice floes, interspersed with open and refrozen leads.

In this study, we assess the accuracy of high resolution CryoSat-2 freeboard measurements, obtained with a novel physical retracker. Whereas other physical models approximate the point target response (PTR) of the radar altimeter (potentially leading to biases in the estimation of range measurements), in our retracker we incorporate the actual across- and along- track PTRs. In addition, we introduce a surface roughness model, that we derived from the Airborne Topographic Mapper (ATM) instrument aboard OIB. The accuracy of the high resolution CryoSat-2 freeboard measurements is assed against ATM freeboard data. Ultimately, we evaluate the performance of delay/Doppler vs high resolution FFSAR for freeboard determination over an extended dataset encompassing a full Arctic season of CryoSat-2 data, from September 2018 until May 2019.

The main difficulties to retrieve sea ice thickness and volume in the southern ocean come from the lack of in-situ observation and knowledge related to sea ice properties. For instance, whereas in-situ observations in the Arctic have enabled to construct snow depth climatologies (e.g the Warren climatology), there are no equivalent snow depth data in the Austral. By consequence, except for a few studies such as Zwally et al, 2008, Kurtz and Markus, 2012 and more recently Kacimi and kwok, 2020, mainly based on based on ICESat's data, no valid sea ice thickness estimations have yet been drawn apart from an experimental ESA-SICCI product (available on the CCI Data Portal, http://cci.esa.int/data), but it does not currently extend beyond the 2016 winter.

In this context, the objective of this presentation is to review the recent developments on sea ice volume estimations in the southern ocean conducted in the framework of an ESA Living Planet Fellowship and the ESA CSAO+ project.
We will first present and compare two radar freeboard solutions calculated from the CryoSat-2 data. The first solution is based on the commonly used TFMRA50 retracker and the second solution has been derived from altimetric ranges calculated on the GPOD platform with the SAMOSA+ physical retracker. Then, we will describe the methodology used to recalibrate the Envisat Low Resolution Mode (LRM) on CryoSat-2 Synthetic Aperture radar (SAR) mode in order to provide homogeneous freeboard estimations.
In a second part, we will evaluate the recent altimetric snow depth product (ASD) computed from the difference of radar penetration between the SARAL/AltiKa Ka-band and the CryoSat-2/SIRAL ku-band radar frequencies (Garnier et al, 2021).
The ASD data will be used to compute the first 100 % altimetric SIT estimation that should also mimic the future CRISTAL mission datasets. This estimation will be compare with the ESA-SICCI product that considers the AMSR passive radiometer snow depth data. In addition, first elements of validation are presented by comparison with some Upward Looking Sonar data (ULS, Behrendt et al, 2013), transects measurements during the Sea Ice Mass Balance in the Antarctic, (SIMBA, Lewis, 2011) and the ICESat's data.
Finally, we will present and analyse tendencies of 2003-2020 sea ice volume time series in the Antarctic.

Southern Ocean sea ice plays an important role in the climate system of the polar regions. In summer it reflects incoming solar radiation creating local cooling, and sea ice formation and melt contribute to Antarctic bottom and intermediate water which help drive the overturning circulation of the Southern Ocean.

Sea ice extent has been routinely monitored by passive microwave satellites since the late 1970’s, and high interannual variability has resulted in little overall trend in the extent of Southern Ocean sea ice cover. However, more recently in 2016 and 2017, there have been dramatic losses of the sea ice cover followed by recovery in 2019 and 2020. Understanding the drivers of these events is critical to gain an insight into whether Antarctic sea ice will transition into a state of decline with ongoing climate change as is predicted by global climate models.

Critical to exploring the drivers of Antarctic sea ice variability is knowing how the total volume of the ice pack has changed, and this requires measurements of its thickness. Sea ice thickness can be measured using satellite altimetry by first measuring the elevation of the sea ice surface relative to the local sea surface height, called the freeboard. From the freeboard, the thickness of sea ice can be estimated assuming sea ice floats in hydrostatic equilibrium in the ocean. This assumption requires an estimate of the thickness of the snow cover that blankets Southern Ocean sea ice throughout the year, as well as the densities of snow, ice and sea water.

In recent years, advances in measuring Southern Ocean snow depth on sea ice have been made using dual-frequency altimetry, but these datasets are limited to the time range when satellites operate simultaneously, e.g. AltiKa + CryoSat-2 (2013 – present), CryoSat-2 + ICESat-2 (2018-present), and so cannot complement the long radar altimetry record that exists since the launch of ERS-1 in the early 1990s.

We present CASSIS, a Lagrangian model which simulates the snow cover on Antarctic sea ice using information about sea ice motion, snowfall, wind redistribution and snow pack processes. Every day, sea ice floes are moved according to ice motion vectors and accumulate snow from the atmosphere and the Antarctic ice sheet. Snow is lost via wind redistribution to leads and transformed into snow-ice. We find that older sea ice has a thicker snow layer, a result which is supported by observations collected by ships. We also analyse trends in snow depth over the 38 years for which the model is run, and find that while the snow cover of Antarctic sea ice has increased overall, it has decreased in the South Pacific Ocean because of a reduction of summer sea ice extent in this region.