Context
Near-real-time accurate Arctic sea ice forecasts of up to 10 days are crucial for many operators in the Arctic. Operational forecasts (e.g. TOPAZ) are based on the integration of observations and the use of a physical-based model to simulate the future evolution of sea ice. Those models contain biases and are expensive to run, limiting the accuracy and the resolution of the delivered product.
Objective of Icecastnet
The IceCastNet project will provide high-resolution sea ice forecasts at a resolution of 1km for a lead time of up to 10 days in the Arctic. Our forecasting system will be based on artificial intelligence techniques and thus will run quickly with low-resource usage enabling the production of an ensemble of forecasts. Our system will use only near-real-time observational products in the production phase. Our approach will constitute the first pan-Arctic sea-ice AI-based short-term forecast in an operational context.
Funding
This work is carried out as part of the Copernicus Marine Service Icecastnet project. Copernicus Marine Service is implemented by Mercator Ocean in the framework of a delegation agreement with the European Union.
Here is the Copernicus Marine Icecastnet project page.
Project management
The Nansen Environmental and Remote Sensing Center in Bergen, Norway is leading the SuperIce project in collaboration with Met Norway