Although snow depth on sea ice is a key parameter for sea ice thickness (SIT) retrieval, there currently does not exist reliable estimations. In the Arctic, nearly all SIT products use a snow depth climatology (the modified Warren-99 climatology, W99m) constructed from in situ data obtained prior to the first significant impacts of climate change. In the Antarctic, the lack of information on snow depth remains a major obstacle in the development of reliable SIT products. In this study, we present the latest version of the altimetric snow depth (ASD) product computed over both hemispheres from the difference of the radar penetration into the snow pack between the Ka-band frequency SARAL/Altika and the Ku-band frequency CryoSat-2. The ASD solution is compared against a wide range of snow depth products including model data (Pan-Arctic Ice-Ocean Modelling and Assimilation System (PIOMAS) or its equivalent in the Antarctic the Global Ice-Ocean Modeling and Assimilation System (GIOMAS), the MERCATOR model, and NASA's Eulerian Snow On Sea Ice Model (NESOSIM, only in the Arctic)), the Advanced Microwave Scanning Radiometer-2 (AMSR2) passive radiometer data, and the Dual-altimeter Snow Thickness (DuST) Ka–Ku product (only in the Arctic). The ASD product is further validated in the Arctic against the ice mass balance (IMB) buoys, the CryoSat Validation Experiment (CryoVEx) and Operation Ice Bridge's (OIB) airborne measurements. These comparisons demonstrate that ASD is a relevant snow depth solution, with spatiotemporal patterns consistent with those of the alternative Ka–Ku DuST product but with a mean bias of about 6.5 cm. We also demonstrate that ASD is consistent with the validation data: comparisons with OIB's airborne snow radar in the Arctic during the period of 2014–2018 show a correlation of 0.66 and a RMSE of about 6 cm. Furthermore, a first-guess monthly climatology has been constructed in the Arctic from the ASD product, which shows a good agreement with OIB during 2009–2012. This climatology is shown to provide a better solution than the W99m climatology when compared with validation data. Finally, we have characterised the SIT uncertainty due to the snow depth from an ensemble of SIT solutions computed for the Arctic by using the different snow depth products previously used in the comparison with the ASD product. During the period of 2013–2019, we found a spatially averaged SIT mean standard deviation of 20 cm. Deviations between SIT estimations due to snow depths can reach up to 77 cm. Using the ASD data instead of W99m to estimate SIT over this time period leads to a reduction in the average SIT of about 30 cm.

Both Arctic and Antarctic sea ice are affected by climate change. While Arctic sea ice has been declining for several decades, Antarctic sea ice extent slowly increased until 2015, followed by a sharp drop in 2016. Quantifying sea ice changes is essential to assess their impacts on the ocean, atmosphere, ecosystems and Arctic communities. In this study, we combine sea ice thickness estimates from four satellite radar altimeters to derive the longest time series of homogeneous sea ice thickness for both hemispheres over 30 years (1994–2023). The record supports the rapid loss of sea ice in the Arctic for each month of the year and the heterogeneous changes in sea ice thickness in the Antarctic. The study confirms that most of the volume variability is due to the thickness variability, which holds true for both hemispheres. The sea ice thickness time series presented here offer new insights for models, in particular the possibility to evaluate sea ice reanalyses and to initialize forecasts, especially in the Antarctic, where the data set presented here has no equivalent in terms of spatial and temporal coverage.

Sea ice volume's significant interannual variability requires long-term series of observations to identify trends in its evolution. Despite improvements in sea ice thickness estimations from altimetry during the past few years thanks to CryoSat-2 and ICESat-2, former ESA radar altimetry missions such as the Environmental Satellite (Envisat) and especially the European Remote-Sensing Satellite (ERS-1 and ERS-2) have remained under-exploited so far. Although solutions have already been proposed to ensure continuity of measurements between CryoSat-2 and Envisat, there is no time series integrating ERS. The purpose of this study is to extend the Arctic radar freeboard time series back to 1995. The difficulty in handling ERS measurements comes from a technical issue known as the pulse blurring effect, altering the radar echoes over sea ice and the resulting surface height estimates. Here we present and apply a correction for this pulse blurring effect. To ensure consistency of the CryoSat-2, Envisat and ERS-2 time series, a multiparameter neural-network-based method to calibrate Envisat against CryoSat-2 and ERS-2 against Envisat is presented. The calibration is trained on the discrepancies observed between the altimeter measurements during the mission-overlap periods and a set of parameters characterizing the sea ice state. Monthly radar freeboards are provided with uncertainty estimations based on a Monte Carlo approach to propagate the uncertainties all along the processing chain, including the neural network. Comparisons of corrected radar freeboards during overlap periods reveal good agreement between the missions, with a mean bias of 0.30 cm and a standard deviation of 9.7 cm for Envisat and CryoSat-2 and a 0.20 cm bias and a standard deviation of 3.8 cm for ERS-2 and Envisat. The monthly corrected radar freeboards obtained from Envisat and ERS-2 are then validated by comparison with several independent datasets such as airborne, mooring, direct-measurement and other altimeter products. Except for two datasets, comparisons lead to correlations ranging from 0.41 to 0.94 for Envisat and from 0.60 to 0.74 for ERS-2. The study finally provides radar freeboard estimation for winters from 1995 to 2021 (from the ERS-2 mission to CryoSat-2).

Antarctic sea ice extent has been persistently low since late 2016, possibly owing to changes in atmospheric and oceanic conditions. However, the relative contributions of the ocean, the atmosphere and the underlying mechanisms by which they have affected sea ice remain uncertain. To investigate possible causes for this sea‐ice decrease, we establish a seasonal timeline of sea ice changes following 2016, using remote sensing observations. Anomalies in the timing of sea ice retreat and advance are examined along with their spatial and interannual relations with various indicators of seasonal sea ice and oceanic changes. They include anomalies in winter ice thickness, spring ice removal rate due to ice melt and transport, and summer sea surface temperature. We find that the ice season has shortened at an unprecedented rate and magnitude, due to earlier retreat and later advance. We attribute this shortening to a winter ice thinning, in line with ice‐albedo feedback processes, with ice transport playing a smaller role. Reduced ice thickness has accelerated spring ice area removal as thinner sea ice requires less time to melt. The consequent earlier sea ice retreat has in turn increased ocean solar heat uptake in summer, ultimately delaying sea ice advance. We speculate that the observed winter sea ice thinning is consistent with previous evidence of subsurface warming of the Southern Ocean.

The relatively stable conditions of the sea ice cover in the Antarctic, observed for almost 40 years, seem to be changing recently. Therefore, it is essential to provide sea ice thickness (SIT) and volume (SIV) estimates in order to anticipate potential multi-scale changes in the Antarctic sea ice. For that purpose, the main objectives of this work are: (1) to assess a new sea ice freeboard, thickness and volume altimetry dataset over 2003–2020 and (2) to identify first order impacts of the sea ice recent conditions. To produce these series, we use a neuronal network to calibrate Envisat radar freeboards onto CryoSat-2 (CS2). This method addresses the impacts of surface roughness on Low Resolution Mode (LRM) measurements. During the 2011 common flight period, we found a mean deviation between Envisat and CryoSat-2 radar freeboards by about 0.5 cm. Using the Advanced Microwave Scanning Radiometer (AMSR) and the dual-frequency Altimetric Snow Depth (ASD) data, our solutions are compared with the Upward looking sonar (ULS) draft data, some in-situ measurement of the SIMBA campaign, the total freeboards of 6 Operation Ice Bridge (OIB) missions and ICESat-2 total freeboards. Over 2003–2020, the global mean radar freeboard decreased by about −14% per decade and the SIT and SIV by about −10% per decade (considering a snow depth climatology). This is marked by a slight increase through 2015, which is directly followed by a strong decrease in 2016. Thereafter, freeboards generally remained low and even continued to decrease in some regions such as the Weddell sea. Considering the 2013–2020 period, for which the ASD data are available, radar freeboards and SIT decreased by about −40% per decade. The SIV decreased by about −60% per decade. After 2016, the low SIT values contrast with the sea ice extent that has rather increased again, reaching near-average values in winter 2020. The regional analysis underlines that such thinning (from 2016) occurs in all regions except the Amundsen-Bellingshausen sea sector. Meanwhile, we observed a reversal of the main regional trends from 2016, which may be the signature of significant ongoing changes in the Antarctic sea ice.

Sea ice volume significant interannual variability requires long-term series of observations to identify trends in its evolution. Despite improvements in sea ice thickness estimations from altimetry during the past few years thanks to CryoSat-2 and ICESat-2, former ESA radar altimetry missions such as Envisat and especially ERS-1 and ERS-2 have remained under-exploited so far. Although solutions have already been proposed to ensure continuity of measurements between CryoSat-2 and Envisat, there is no time series integrating ERS. The purpose of this study is to extend the Arctic freeboard time series back to 1995. The difficulty to handle ERS measurements comes from a technical issue known as the pulse-blurring effect, altering the radar echos over sea ice and the resulting surface height estimates. Here we present and apply a correction for this pulse-blurring effect. To ensure consistency of the CryoSat-2/Envisat/ERS-2 time series, a multi-parameters neural network-based method to calibrate Envisat against CryoSat-2 and ERS-2 against Envisat is presented. The calibration is trained on the discrepancies observed between the altimeter measurements during the missions-overlap periods and a set of parameters characterizing the sea ice state. Monthly radar freeboards are provided with uncertainty estimations based on a Monte Carlo approach to propagate the uncertainties all along the processing chain, including the neural network. Comparisons of corrected radar freeboards during overlap periods reveal good consistencies between missions, with a mean bias of 3 mm for Envisat/CryoSat-2 and 2 mm for ERS-2/Envisat. The monthly maps obtained from Envisat and ERS-2 are then validated by comparison with several independent data such as airborne, moorings, direct measurements and other altimeter products. Except for two data sets, comparisons lead to correlation ranging from 0.42 to 0.94 for Envisat, and 0.6 to 0.76 for ERS-2. The study finally provides radar freeboard estimation for winters from 1995 to 2021 (from ERS-2 mission to CryoSat-2).