Sea ice is governed by a variety of small-scale processes that affect its large-scale evolution. Modelling this nonlinear coupled multidimensional system remains a major challenge, because (1) we still lack the understanding of the physics governing sea-ice dynamics and thermodynamics, (2) observations to conduct model evaluation are scarce and (3) the numerical approximation and the simulation become more difficult and computationally expensive at higher resolution.
Recently, several new modelling approaches have been developed and refined to address these issues. These include but are not limited to new rheologies, discrete element models, advanced subgrid parameterizations, the representation of wave-ice interactions, sophisticated data assimilation schemes, often with the integration of machine learning techniques. Moreover, novel in-situ observations and the growing availability and quality of sea-ice remote-sensing data bring new opportunities for improving sea-ice models.
This session aims to bring together researchers working on the development of sea-ice models, from small to large scales and for a wide range of applications such as idealised experiments, operational predictions, or climate simulations, to discuss current advances and challenges ahead.
Sea ice remains challenging to predict across spatial and temporal scales, from hourly floe-scale motion to seasonal regional forecasts and long-term climate projections. Increasingly complex numerical models have been developed to represent the strongly nonlinear dynamics and thermodynamics of sea ice, but their growing computational cost together with uncertainties in initial conditions and unresolved physical processes continues to limit their applicability. In recent years, machine-learning approaches trained on satellite observations or numerical model output have shown promise in predicting sea-ice evolution. However, while often accurate, such purely data-driven models provide limited physical interpretability, hindering the analysis of individual dynamic and thermodynamic processes and their response to climate change. Here we present a hybrid modelling framework that bridges these two approaches: a machine learning-enabled numerical sea-ice model. At its core is a differentiable implementation of a dynamic-thermodynamic sea-ice model in Python, which allows the computation of sensitivities with respect to model parameters as well as initial and boundary conditions. This enables systematic parameter optimization against observations and, more importantly, facilitates the replacement of individual parameterizations with lightweight machine-learning components. Once trained, these lightweight components remain physically interpretable due to their low complexity, explicit input-output relationships, and strictly local (pointwise) operation, in contrast to black-box, high-dimensional ML models. The hybrid model is embedded in an efficient data-loading infrastructure that provides access to diverse observational data sources, including satellite, buoy, and in-situ measurements, for training and evaluation. We demonstrate the capabilities of this framework by comparing alternative data-driven thermodynamic parameterizations for sea-ice and snow growth and melting rates trained on buoy and satellite data, and by assessing their impact on large-scale sea-ice evolution.
How to cite: Hutter, N., Wilms, A., and Juricke, S.: Learning sea-ice physics from data: a hybrid ML-numerical modelling framework, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9145, https://doi.org/10.5194/egusphere-egu26-9145, 2026.
We present an efficient hybrid Neural Network–Finite Element Method (NN-FEM) for the viscous–plastic (VP) sea-ice model used in climate simulations. VP solvers are costly due to the strongly nonlinear material law, with cost per degree of freedom increasing rapidly under mesh refinement. However, high resolution is needed to capture narrow deformation bands (linear kinematic features). Our approach enriches coarse-mesh FEM solutions with fine-scale corrections predicted by a locally applied neural network trained on high-resolution data. The patch-based network is small, parallelizable, and generalizes across right-hand sides and domains. Numerically, the method achieves comparable accuracy at approximately 11× lower cost and speeds Newton iterations by up to 10%.
How to cite: Margenberg, N. and Mehlmann, C.: A Hybrid Neural Network-Finite Element Method for the Viscous-Plastic Sea-Ice Model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4488, https://doi.org/10.5194/egusphere-egu26-4488, 2026.
Antarctic sea ice plays a critical role in modulating global climate variability and in supporting polar ecosystems. However, sea ice seasonal forecasts continue to show limited skill in capturing interannual variability across key regions and seasons. Here we evaluate seasonal predictions of Antarctic sea ice extent predictions from six Copernicus Climate Change Service forecast systems, focusing on the Weddell Sea, where springtime skill deteriorates rapidly with lead time. We identify two systematic sources of model error. First, models show excessive persistence of winter sea ice anomalies compared to observations, indicating an overdependence on ocean conditions. Second, they fail to adequately represent large-scale atmospheric circulation anomalies associated with El Niño–Southern Oscillation and Southern Annular Mode interactions, underestimating Amundsen Sea Low pressure anomalies and related wind patterns. These circulation-related biases appear to originate from misrepresented atmospheric responses to tropical Pacific sea surface temperatures. Our results prompt us to revisit sea-ice ocean couplings and better capture tropical-Antarctic teleconnections in dynamic models to improve Antarctic sea ice prediction.
How to cite: Gao, Y., Nie, Y., Massonnet, F., Xiu, Y., Topál, D., Luo, H., Lv, X., and Yang, Q.: Biases in tropical Pacific teleconnections and ocean memory together limit spring sea ice predictability in dynamical models in the Weddell Sea, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2833, https://doi.org/10.5194/egusphere-egu26-2833, 2026.
The Met Office GC5 climate model has a simple sea ice albedo. To improve this, new albedo schemes have been implemented for bare sea ice, snow on sea ice and melt ponds. These improved albedo schemes include a solar zenith angle dependence, a snow grain size dependence (for snow on sea ice) and a more scientifically accurate representation of absorption of sunlight within melt ponds.
Climate simulations of GC5 for present day and 2xCO2 have been performed with (and without) the improved sea ice albedo schemes. We will present the effect of improved sea ice albedos on the present climate and the effect when a doubling of CO2 has been imposed.
How to cite: Copsey, D.: New sea ice albedo schemes to be included in future Met Office climate models and their effect on the climate., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10773, https://doi.org/10.5194/egusphere-egu26-10773, 2026.
The horizontal size of a sea ice floe directly impacts both climatic feedbacks and sea ice dynamics. The floe size distribution (FSD) quantifies how the number of sea ice floes varies with floe size across a domain. The creation of models that can reproduce an accurate representation of the FSD is paramount for both climate prediction and seasonal Arctic forecasts, as floe size contributes to the ice-albedo feedback and sea ice evolution. Within the Community Ice CodE (CICE) sea ice model, sea ice floes are modelled using a prognostic FSD, where five processes may alter the FSD: lateral growth, lateral melt, new ice creation, welding (when two different floes coagulate), and wave fracture. Field notes have observed that when two floes weld together, the two floes may be bridged with younger ice providing an intermediary. We propose that the younger ice will be weaker, making the newly welded floe more likely to fracture. Yet CICE’s prognostic FSD carries no memory of prior welding or interactions, preventing the model from capturing this history-dependent weakness and how it influences dynamical processes, such as wave fracture or floe deformation. To account for this, we introduce a new parameterization for a floe’s health that calculates the fraction of floes in a given grid cell that have recently welded and that will be most susceptible to dynamic fragmentation. The parameterization also accounts for the effects of floes healing over time (with the welded bridge thickening), the effects of dynamic fracture on floe health, and the advection of the floe health metric through the model’s domain. With our accounting of floe health, we create a linear floe health dependent feedback that strengthens the influence of wave fracture and floe deformation. Finally, we use the new floe health metric to couple the FSD to CICE’s ice strength parameterization, weakening the ice in grid cells where many floes are the product of recent welding.
How to cite: Webb, G. and Bitz, C. M.: A new parameterization describing the health of a sea ice floe and its potential influence on wave fracture, ice deformation, and ice motion , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14808, https://doi.org/10.5194/egusphere-egu26-14808, 2026.
Sea ice is a highly heterogeneous, granular material whose mechanical behavior arises from interactions among individual floes. This granular nature makes discrete element methods (DEMs) a natural choice for modeling sea ice dynamics. In principle, DEMs can resolve discontinuities, such as fractures and leads, and directly capture deformation, ridging, and dilatation processes without relying on parameterizations. However, several challenges (e.g., contact detection, floe geometry, open water treatment) limit their applicability to large-scale simulations. For these reasons, computational efficiency, and historical precedent, continuum models have long been the community’s choice for simulating sea ice at larger scales, even though they face their own challenges. Although running high-resolution (<2 km) state-of-the-art continuum models improves the representation of linear kinematic features (LKFs) and deformation statistics, this is computationally expensive, and Earth system models running at coarser resolutions can’t benefit from this. Thus, they are supplemented with parameterizations to account for subgrid-scale processes that cannot be captured by rheological models alone. Therefore, a modeling framework that bridges the gap between particle-resolving and continuum scales is required.
We present a new sea ice discrete element model, the Granular flOe Dynamics for seA ice Rheology (GODAR) model, to bridge the gap between the engineering and pan-Arctic scales. The purpose of this model is to support the development of parameterizations representing the granular behavior of sea ice in continuum sea ice models. GODAR includes a ridging parameterization, a rolling-resistance model that captures complex geometries, and a novel sheltering parameterization. The model explicitly resolves the formation and evolution of LKFs and reproduces the dilatant behavior observed in granular materials, while maintaining computational tractability suitable for regional domains by using cylindrical particles. Results from shear experiments demonstrate that ridges are highly localized along the failure planes and that the ridge build-up coincides with a positive angle of dilatancy (compressive regime). The novel geometric sheltering coefficient can reduce the total form drag of the sheltered floes, resulting in a non-zero moment on assemblies of floes. Finally, GODAR could be easily improved to run basin-scale simulations in the near future, and be extended to a seamless model capable of representing all ice types – sea ice, icebergs, ice shelves, and land ice – within a unified framework.
How to cite: Savard, A., Tremblay, B., and Polojärvi, A.: GODAR: a new discrete element sea ice model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7618, https://doi.org/10.5194/egusphere-egu26-7618, 2026.
How to cite: Arlen, A., Wilson, E., Dansereau, V., Meng, Y. (., and Lai, C.-Y.: A surprisingly capable minimal bonded discrete element model for sea ice, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13808, https://doi.org/10.5194/egusphere-egu26-13808, 2026.
We present IceCastNet, a data-driven model for forecasting Arctic sea ice state (including concentration, thickness, and drift) up to 10 days ahead. The model is trained on satellite-derived products: OSI-SAF for sea ice concentration and drift, and CS2SMOS for sea ice thickness. During the melt season, when CS2SMOS data are unavailable, the dataset is supplemented with the TOPAZ reanalysis. IceCastNet also uses meteorological forecasts (10-meter wind and 2-meter air temperature) from ECMWF as forcing.
The architecture is based on a graph-transformer design, similar to that used in the Artificial Intelligence/Integrated Forecasting System (AIFS), and implemented within ECMWF’s Anemoi framework. IceCastNet achieves skills comparable to, and in some cases better than, established baselines such as the operational TOPAZ system, particularly for sea ice concentration. These improvements appear to stem from reduced biases in initial conditions and lower forecast error. This is assessed by comparing IceCastNet outputs with debiased TOPAZ forecasts and independent sea ice concentration products, including SAR-derived estimates from the Danish Meteorological Institute (ASIP) and ice charts produced by experts at the U.S. National Ice Center.
The inference time for a 10-day forecast with IceCastNet of about 10 seconds is approximately two orders of magnitude shorter than that of physics-based systems such as TOPAZ. This substantial reduction in computational cost makes IceCastNet a computationally efficient alternative, although IceCastNet only provides the observed variables. Moreover, since it relies exclusively on operational, near-real-time data, IceCastNet is well-suited for integration into operational sea ice forecasting workflows.
The spatial resolution of IceCastNet forecasts follows that of the training data. We also show that applying a super-resolution procedure trained on high-resolution sea ice simulations from the model neXtSIM can enhance the resolution of IceCastNet outputs.
References:
CS2SMOS: https://doi.org/10.5194/tc-11-1607-2017
OSI-SAF: https://osi-saf.eumetsat.int/
AIFS: https://arxiv.org/abs/2406.01465
TOPAZ: https://doi.org/10.48670/moi-00001
ASIP: https://ocean.dmi.dk/asip/
neXtSIM: https://egusphere.copernicus.org/preprints/2025/egusphere-2024-3521/
How to cite: Brajard, J., Edel, L., Palerme, C., Korosov, A., and Bertino, L.: End-to-end forecast of the Arctic sea ice initialised directly from observations, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6436, https://doi.org/10.5194/egusphere-egu26-6436, 2026.
Large-scale sea ice models still struggle to reproduce observed drift and deformation patterns, underscoring the need for physically based representations of unresolved floe-scale processes. Advances in high-resolution modelling and satellite imagery now offer new opportunities to investigate floe-scale dynamics and their influence on large-scale ice behaviour. These developments are essential not only for improving model fidelity but also for Arctic navigation and offshore safety.
As part of the ESA-funded HIRLOMAP project, we present a workflow for generating high-resolution sea ice products by integrating satellite observations, image processing, and numerical modelling. The study domain is a 110 × 110 km southeast of Svalbard. Two Sentinel-2 images (10 m resolution) acquired on 16th and 17th of April 2025 are used. Ice drift vectors were derived from the satellite images through feature tracking, enabling estimation of local displacement and deformation rates (divergence and shear).
To simulate ice dynamics, we employ a Discrete Element Model (DEM) with boundary conditions informed by satellite-derived drift. Environmental forcing includes wind fields from NORA3 and ocean currents from Barents-2.5, while ice thickness is obtained from CryoSat-2/SMOS products. The initial ice field is digitized from Sentinel-2 imagery, yielding around 40 000 floes. Computational efficiency is improved through hierarchical clustering and area-based filtering, reducing floe count to around 1 800 while conserving total ice concentration. Model calibration focuses on air drag coefficients to reproduce observed deformation patterns. Simulated drift and strain rates are compared against Sentinel-2 observations and Barents-2.5 outputs, demonstrating the capability of DEM to capture local-scale variability beyond continuum models.
Future work will address large-scale fracture processes (e.g., ridging), wave–ice interactions, strategies to enhance computational performance, and the integration of machine learning approaches to further advance modelling capabilities. Even though these processes are not yet included, the results presented here already demonstrate strong potential for delivering next-generation Arctic sea ice services that combine high-resolution satellite data and DEM-based modelling. Beyond its engineering applications, this approach demonstrates how DEM can act as a subgrid parameterization tool for continuum models, enabling large-scale systems to represent floe-scale processes such as deformation and fracture.
How to cite: Lubbad, R., Yang, B., Hughes, N., Lu, W., Løset, S., and Dierking, W.: High-Resolution Sea Ice Drift Modelling Using Sentinel-2 and Discrete Element Method: Towards Subgrid Parameterization for Large-Scale Models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10536, https://doi.org/10.5194/egusphere-egu26-10536, 2026.
Sea-ice salt dynamics control ice structure and ice–ocean salt fluxes, yet they remain weakly constrained at large scale and poorly evaluated in models. Here we compile a new global database of >1,200 sea-ice core salinity profiles and use it to provide the first large-scale evaluation of global 1° NEMO-SI3 simulations with new options for salt dynamics, including both state-of-the-art and historical reference parameterizations. The observations reveal a robust Arctic–Antarctic contrast: Antarctic sea ice is on average saltier and more variable, with differences expressed in the seasonal cycle, thickness space, and vertical structure, consistent with more frequent flooding and weaker flushing. Using this dataset, we systematically evaluate key model choices controlling salt dynamics, including prognostic versus diagnostic vertical salinity, gravity-drainage parameterizations, and the new-ice liquid fraction. With appropriate parameter settings, the model reproduces observed hemispheric contrasts, vertical profiles, and salinity–thickness relationships, while highlighting larger uncertainties and process gaps in Antarctic salt dynamics.
How to cite: Vancoppenolle, M., Ortega, E., Rousset, C., Lemaire, E., and Moreau, S.: A large-scale evaluation of sea-ice salt dynamics using new observations and NEMO-SI3 simulations, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-22027, https://doi.org/10.5194/egusphere-egu26-22027, 2026.
NeXtSIM-DG is a novel sea ice model developed as part of the Scale Aware Sea Ice Project (SASIP).
The continum model supports both the established viscous plastic rheology (mEVP) and the Brittle Bingham-Maxwell rheology (BBM). The discretization is based on higher-order discontinuous and continuous finite elements to accurately represent the sharp features of sea ice. Quadrilateral parametric meshes with regular topology are used to allow for easy coupling with ocean models and a highly efficient implementation.
Following best practices in software engineering, the C++ code is designed to be maintainable and easily extendable, with a modular design that allows users to add new physics implementations.
In this poster, we give an overview of neXtSIM-DG and present recent developments regarding its usability and performance. We demonstrate our model's heterogeneous compute capabilities, supporting shared-memory parallelization with OpenMP, full GPU acceleration through Kokkos and scaling across multiple CPUs via MPI.
How to cite: Jendersie, R., Ólason, E., Spain, T., Richter, T., Meltzer, T., Wallwork, J., Albert, A., and Shah, N. V.: neXtSIM-DG: The next-generation discontinuous Galerkin sea ice model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11333, https://doi.org/10.5194/egusphere-egu26-11333, 2026.
Arctic landfast ice (LFI) is sea ice that remains mechanically immobilized along Arctic coastlines for prolonged periods. LFI influences the stability of the Arctic halocline by displacing coastal polynyas offshore and modifying the mixing of river plumes. In large-scale ocean–sea ice models, a realistic representation of Arctic LFI relies on two interacting modelling components: a grounding (basal stress) parameterization, which establishes anchor points in shallow waters, and the sea ice rheology, which controls ice deformation between these anchors. Here we examine the sensitivity of simulated Arctic LFI to these components and to their interaction. Simulations are performed using the Nucleus for European Modelling of the Ocean–Sea Ice Modelling Integrated Initiative (NEMO-SI³) platform on a 0.25° global grid. Two contrasting rheological formulations are considered: the adaptive elastic-viscous-plastic rheology with tensile strength (aEVPts) and the brittle Bingham-Maxwell rheology (BBM), each tested with and without a grounding scheme. Owing to its elastic component, BBM exhibits a more rigid mechanical behaviour than aEVPts and more readily immobilizes sea ice between anchor points, resulting in enhanced LFI formation. However, this increased rigidity of BBM also limits ice thickening in convergence zones, thereby reducing the effectiveness of the grounding scheme through a decrease in the number of anchor points. Comparison with in situ observations of Arctic LFI thickness further highlights the importance of accurately representing the timing and duration of sea ice immobilization. When sea ice becomes immobilized early and subsequently grows predominantly through thermodynamic processes, model biases in LFI thickness are significantly reduced, consistent with Stefan’s law for thermodynamic ice growth.
How to cite: Lambotte, A., Fichefet, T., Massonnet, F., Brodeau, L., Rampal, P., Lemieux, J.-F., Dupont, F., and Vancoppenolle, M.: Distinct and synergistic influences of sea ice rheology and basal stress on the simulation of Arctic landfast ice, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13276, https://doi.org/10.5194/egusphere-egu26-13276, 2026.
When ocean surface waves are present in ice-covered seas, they can cause sea ice to bend and fracture. Wave-induced fracture of sea ice could play an increasingly important role in the evolution of the Arctic and Antarctic sea ice covers. We present coupled simulations using the development version of the Community Earth System Model (CESM3) that now includes interactions between waves and sea ice, coupled via the sea ice floe size distribution. In particular, we show the influence of a unified sea ice fracture model based on elastic beam theory for the bending of a sea ice plate. This fracture model is valid for all wavelengths and spans the fully flexible and fully rigid limits, unlike previous approaches. Early results suggest that this model improves simulation of sea ice fracture and the floe size distribution, with implications for sea ice feedbacks. Further, we demonstrate a machine-learning-based emulation of the fracture scheme to reduce computational expense.
How to cite: Roach, L. A., Bailey, D. A., Bitz, C. M., DuVivier, A., Holland, M. M., Tremblay, B., and Webb, G.: Fracture of sea ice by ocean surface waves in CESM3, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16184, https://doi.org/10.5194/egusphere-egu26-16184, 2026.
The accurate simulation of sea ice deformation and fracturing remains a significant challenge, driving the development of advanced continuum rheologies. In particular, brittle (elasto-brittle or EB, Maxwell elasto-brittle or MEB, brittle Bingham-Maxwell or BBM) rheologies are a promising advancement, introducing a damage parameter that accounts for sub-grid scale fractures and material degradation under high stresses without requiring large deformations. They simulate realistic large scale fields with adequate heterogeneity and intermittency even at coarser resolution.
However, the conventional MEB implementation, which relies on correcting super-critical stresses to bind the simulated stress to the yield criterion, inadvertently introduces a growth of numerical errors in the stress field. These errors can be reduced by introducing the Plante and Tremblay (2021) generalized stress correction scheme.
Here, the generalized stress correction scheme is added to the implemenation of the MEB rheology in the sea ice component of the Massachussetts Institute of Technology general circulation model (MITgcm), a community model that also includes a viscous-plastic (VP) rheology.
The results of pan-Arctic simulations with different levels of numerical errors in the stress field are compared against simulations using the VP rheology to identify the effect of numerical errors on the deformation behaviour. Our findings reveal three critical insights. First, we find that the generalized correction scheme successfully reduces numerical errors in the stress field in the complex Arctic simulations. Second, we show that reducing the numerical errors effectively reduce the number of simulated Linear Kinematic Features (LKFs). This confirms that numerical errors are partly responsible for the generated spatial heterogeneity in the MEB rheology. Third, we introduce artificial numerical errors to the yield curve of both MEB and VP to find that the MEB rheology is a lot more sensitive to it than the VP rheology. Additionally, we use idealized scenarios to isolate the problem from the complexitiy of an Arctic simulation. In the idealized experiments we can reproduce our results and find that the fracture angle seems to be dependent on the generalized correction scheme, as well.
Our work leads to more understanding of the MEB rheology with the goal of finding the source of heterogeneity and the seeding of LKFs in sea ice models.
How to cite: Bourgett, M., Losch, M., and Plante, M.: Implications of Numerical Error Growth in the Maxwell Elasto-Brittle Rheology for Sea Ice Fracturing and Spatial Heterogeneity, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-483, https://doi.org/10.5194/egusphere-egu26-483, 2026.
Sea ice dynamics in numerical models are discretized on a spatial grid, with variables located at grid cell vertices, edges, or centers. On triangular grids, the number of edges is three times larger than the number of vertices. Consequently, placing variables on edges --- corresponding to a non-conforming finite-element discretization --- increases the number of degrees of freedom and effectively enhances the spatial resolution on a given mesh. Idealized benchmark experiments without a coupled ocean using the CD-grid discretization, where sea ice velocity is placed on edges and tracers on vertices, have demonstrated an increased occurrence of small-scale fractures and more finely resolved features in the sea ice field compared to other discretizations. In this study, we present the first application of the CD-grid in a high-resolution, fully coupled sea ice--ocean simulation using FESOM2, using a spatial resolution in the Arctic of 4.5 km. We show that the CD-grid produces a more sharply resolved sea ice field with an increased occurrence of small-scale structures relative to the A-grid configuration, in which both velocity and tracers are placed on vertices, while maintaining the same large-scale sea ice characteristics. In addition, we present a scaling analysis of the CD-grid in parallel applications across varying core counts, including runtime benchmarks and a direct performance comparison with the A-grid.
How to cite: Gärtner, J. and Danilov, S.: Sharper Resolution of Arctic Sea Ice Dynamics with Non-Conforming Finite Elements in FESOM2, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-438, https://doi.org/10.5194/egusphere-egu26-438, 2026.
Observations since 1990s reveal widespread warming in the deep and bottom Arctic Ocean. It is historically attributed to geothermal heating, whereas the impacts of global and Arctic climate change on the deep and bottom Arctic Ocean warming remain unresolved. Our study demonstrates that during the recent decades, the Arctic Ocean deep water is warming at 0.020 ℃/decade in the Eurasian Basin between 2000 and 2600 m, exceeding what can be explained by geothermal heating. We find that the rapid warming in the deep Greenland Basin diminishes its cooling effect on the deep Eurasian Basin via the Fram Strait, leading to the warming in the deep Eurasian Basin. Meanwhile, the Lomonosov Ridge blocks this warming signal from reaching the deep Amerasian Basin, maintaining its relative slow warming rate of 0.003 ℃/decade. Our findings indicate that the deep Greenland Basin warming has already exerted obvious impacts on the deep Arctic Ocean.
How to cite: Song, R., Chen, X., Bi, H., Wang, X., and Mu, L.: Deep Arctic Ocean Warming Enhanced by Heat Transferred from Deep Atlantic, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-599, https://doi.org/10.5194/egusphere-egu26-599, 2026.
