15 September 2023 in Ocean Science
In the last decade, various satellite missions have been monitoring the status of the cryosphere and its evolution. Besides sea-ice concentration data, available since the 1980s, sea-ice thickness retrievals are now ready to be used in global operational prediction and global reanalysis systems. Nevertheless, while univariate algorithms are commonly used to constrain sea-ice area or volume, multivariate approaches have not yet been employed due to the highly non-Gaussian distribution of sea-ice variables together with the low accuracy of thickness observations. This study extends a 3DVar system, called OceanVar, which is routinely employed in the production of global/regional operational/reanalysis products, to process sea-ice variables. The tangent/adjoint versions of an anamorphosis operator are used to locally transform the sea-ice anomalies into Gaussian control variables and back, minimizing in the latter space. The benefit achieved by such a transformation is described. Several sensitivity experiments are carried out using a suite of diverse datasets. The sole assimilation of the CryoSat-2 provides a good spatial representation of thickness distribution but still overestimates the total volume that requires the inclusion of Soil Moisture and Ocean Salinity (SMOS) mission data to converge towards the observation estimates. The intermittent availability of thickness data can lead to potential jumps in the evolution of the volume and requires a dedicated tuning. The use of the merged L4 product CS2SMOS shows the best skill score when validated against independent measurements during the melting season when satellite data are not available. This new sea-ice module is meant to simplify the future coupling with ocean variables.
5 February 2026 in EarthArXiv
Biogeochemical models form a core part of marine forecasting and climate projections, yet they suffer from persistent biases in predicting key ecosystem variables, creating challenges across regional and global scales. To address this, we developed an AI-augmented three-dimensional hybrid framework that integrates machine-learning corrections directly into a process-based model’s productivity engine at runtime, keeping mechanistic formulations central while deploying physics-constrained, data-driven AI adjustments around them. We explored two independent hybrid pathways: a satellite-trained primary production scale-factor and a physiology-informed parameter adjustment. Using a temperate shelf-sea as a testbed, we evaluated multi-year hybrid simulations against in situ, Argo, and satellite observations, as well as data-assimilative (DA) reanalysis and high-resolution simulations. Results show that the hybrid framework substantially reduced long-standing biases and outperformed reanalysis and high-resolution simulations across several metrics, including evaluation years and variables, not seen during AI training. This demonstrates that correcting ecosystem process representations while remaining mass-conservative can yield greater accuracy than increasing spatial resolution or relying entirely on continuous DA. Furthermore, because our AI components utilise globally available satellite and experimental datasets, our framework is potentially transferable across global shelf seas. This low-computational, interpretable approach could deliver an effective alternative for operational forecasting and long-term climate applications.
22 February 2023 in EGUsphere
In the last decade, various satellite missions have been monitoring the status of cryoshopere and its evolution over time. Beside sea-ice concentration data, available since the 80s, sea-ice thickness retrievals are now ready to be used in operational prediction and reanalysis systems. Nevertheless, a straightforward ingestion of multiple sea-ice characteristics in a multivariate framework is prevented by the highly non-gaussian distribution of such variables together with the low accuracy of thickness observations. This study describes an extension of OceanVar, a 3Dvar system routinely employed in the production of global/regional operational/reanalysis products, designed to include sea-ice variables. Those variables are treated through an anamorphosis operator that transforms sea-ice anomalies into gaussian control variables, the benefit brought by such transformation is described. Several sensitivity experiments are carried out using a suite of diverse datasets. The assimilation of the sole Cryosat-2 provides a good spatial representation of thickness distribution but still overestimates the total volume that requires the inclusion of SMOS data to be properly constrained. The intermittent availability of thickness data along the year, leads to potential discontinuities in the integrated quantities that requires a dedicated tuning. The use of merged L4 product CS2SMOS produces similar skill score when validated against independent mooring data, compared to the ingestion of L3 CryoSat-2 and L3 SMOS data. The new sea-ice module is meant to simplify the future coupling with ocean variables.