Sara Fleury (LEGOS, Université de Toulouse, CNES, CNRS, IRD, UPS, Toulouse, France); Pierre Rampal (Université Grenoble Alpes, CNRS, IRD, IGE, Grenoble, France); Gwenaël Jestin (LEGOS, France); Florian Le Guillou (Datlas, France); Clément Ubelmann (Datlas, France)

Observing sea ice concentration (SIC) and thickness (SIT) remains a key challenge for climate monitoring and modeling. Satellite altimetry provides measurements of ice and sea level anomaly from which can be estimated SIC, sea ice freeboard, and ultimately SIT. An accurate and well-resolved gridded estimation of these quantities would significantly improve our ability to quantify sea ice volume and its variability, support short- to medium-term forecasting of sea ice conditions and, more generally, improve our understanding of sea ice dynamics. But, as it is often the case, the mapping process is hampered by the sparsity of valid observations. Traditional methods based on monthly gridding or averaging introduce biases in dynamical regions and fail to capture sharp SIT gradients, especially near the ice edge or in ridged ice areas.
In this study, we develop a novel reconstruction approach: VarDyn, based on a spatio-temporal basis decomposition, inspired by recent advances in sea surface height mapping. The method expresses SIC or SIT as a linear combination of basis functions defined over multiple spatial and temporal scales. Several families of basis (e.g., Gaussian, Haar, wavelets) and combinations thereof are tested. A variational framework is then used to constrain the SIC/SIT reconstruction to best fit available SWOT observations, while promoting physically plausible structures. Preliminary results in a synthetic experiment setting over the PanArctic region demonstrate improved representation of both large-scale SIT gradients and localized features such as ridges. Against model truth, VarDyn performs well compared to standard averaging approaches in terms of RMS error and outperforms them in terms of spatial and temporal spectral scores.
In addition, we present the first daily gridded SIC maps using the VarDyn framework and derived from the SWOT lead/floe Level 3 classification. The use of SWOT’s high-resolution observations enables the detection of leads and floes with unprecedented detail. The resulting gridded reconstruction hence captures fine-scale structures of the sea ice cover and provides temporally consistent information at high resolution.
A promising avenue for future development is the integration of sea ice drift as a dynamical constraint within the VarDyn variational formulation. Incorporating this constraint would enhance the temporal coherence of the gridded reconstructions. Overall, this work opens new perspectives for generating SWOT-based high-resolution SIC and SIT Level 4 products.