Hybrid modelling with deep learning for improved sea-ice forecasting

• Main Authors: Finn, T., Durand, C., Farchi, A., Bocquet, M., Chen, Y., Carassi, A., Dansereau, V., Ólason, E.
• Format: Conference Object
• Language: English
• Published: 2023
• Subjects: Sea ice
• Online Access: https://gfzpublic.gfz-potsdam.de/pubman/item/item_5019666

We present our vision on how to advance short-term sea-ice forecasting with deep learning, based on two specific examples. To incorporate multifractal, anisotropic, and stochastic-like processes in sea ice, we envision the combination of geophysical sea-ice models together with neural networks in a hybrid modelling setup. On the one hand, deep learning can surrogate computationally expensive sea-ice models, like neXtSIM. This not only allows us to speed-up simulations by orders of magnitude, but also to improve forecasts of sea-ice thickness by up to 35 % compared to persistence on a daily timescale. On the other hand, deep learning can parametrize subgrid-scale processes in sea-ice models and correct persisting model errors, improving the forecasts by up to 70 % across all model variables on an hourly timescale. Based on these results, we conclude that hybrid modelling with deep learning can lead to major advancements in sea-ice forecasting.