ICEBOX project

Sea ice forecasts -in a box- for informed Arctic Navigation

IceBox concept figure

The concept

IceBox aims to transform Arctic navigation by bringing powerful, AI-driven forecasting directly onto ships. The project will develop a portable “Forecast-in-a-Box” system, a compact computing platform equipped with modern graphics processors that can collect data, run advanced machine learning models, and deliver ultra-high-resolution sea-ice and weather forecasts in real time. By combining satellite imagery, drifting buoy measurements, and onboard sensors, IceBox will generate detailed forecasts with resolutions down to around 100 meters, tailored to the vessel’s immediate surroundings. A key innovation is running the entire forecasting workflow locally or through edge computing, allowing ships to produce on-demand predictions even in remote regions with limited connectivity. The system will be co-designed and tested in close collaboration with users, ensuring that it meets real operational needs in harsh Arctic environments. By making high-resolution, real-time forecasts directly accessible to those navigating the Arctic, IceBox will improve safety, reduce environmental risks, and support more efficient maritime operations. The project will also share data, tools, and results openly, contributing to scientific progress and the development of next-generation digital services for the Arctic.

24-hour AI-generated sequence north of Svalbard
The animation shows a 24-hour AI-generated sequence based on two input Landsat-8 patches north of Svalbard. The method used is the Real-time Intermediate Flow Estimation algorithm (RIFE, Huang et al. 2020). The colored dots indicate positions of buoys deployed during the SvalMIZ-24 campaign.

The project team

Norwegian Meteorological Institute, Norway

Dr. Malte Muller

Dr. Malte Müller is the lead-PI of the IceBox project. Malte is a senior scientist at the Norwegian Meteorological Institute. His research interests include high-resolution coupled ocean, sea-ice, and atmosphere modelling and the provision of weather and climate information to end-users.

Dr. Jean Rabault

Dr. Jean Rabault co-leads Work Package 1. Jean is a senior engineer at the Norwegian Meteorological Institute. His research is at the interface between hardware development, machine learning, and oceanography. In particular, Jean has developed the OpenMetBuoy and been involved in collecting large dataset of sea ice drift, waves in ice, and temperature in the Arctic which are used to further develop and validate sea ice models.

Dr. Paulina Tedesco

Dr. Paulina Tedesco leads Work Package 2. Paulina is a researcher at the Norwegian Meteorological Institute. Her research focuses on applying machine learning techniques for weather and ocean modelling, particularly in the Arctic.

Havard Futsater

Håvard Futsæter is PI and leads Work Package 3. Håvard is at the Norwegian Meteorological Institute (MET) and is an expert in IT architecture development, and co-PI in the Forecast-in-the-Box DestinE project.

Dr. Dina Abdel-Fattah

Dr. Dina Abdel-Fattah leads Work Package 4. Dina is a researcher at the Norwegian Meteorological Institute and the manager of the International Coordination Office of the WMO-WWRP Polar Coupled Analysis and Prediction for Services (PCAPS) project. Her research interests include optimising the co-production process of scientific information products such that they reflect and support different user needs, particularly in a polar and cryospheric hazard context.

Nansen Environmental and Remote Sensing Center, Norway

Dr. Anton Korosov

Dr. Anton Korosov leads Work Package 1. Anton Korosov is a senior researcher at NERSC with expertise in SAR and passive microwave observations of sea ice properties, including sea ice type, sea ice drift and deformation, and sea ice age. He develops assimilation of sea ice concentration and damage into the Lagrangian sea ice model (neXtSIM) and a platform for operational forecasting of sea ice.

Dr. Julian Brajard

Dr. Julien Brajard is co-PI and co-leads Work Package 2. Julien is a researcher at NERSC and an expert in machine learning-based sea-ice forecasting, currently leading projects with Copernicus Marine Service (CMS) and the European Space Agency (ESA).

Henrike Wilborn

Henrike Wilborn leads Work Package 5. She is a communication advisor at NERSC and has a background in geosciences.