Images of climate innovation
IceNet:Forecasting Arctic sea ice with AI
The area covered by sea ice in the Arctic has halved over the past four decades - a reduction equal to 25 times the size of Great Britain. This has dramatic consequences for polar ecosystems and indigenous peoples who depend on sea ice for food and travel. Researchers at the British Antarctic Survey have designed a pioneering new sea ice forecasting artificial intelligence called 'IceNet', which could anticipate sea ice loss before it happens.
The Arctic is a region on the frontline of climate change, warming at 3 times the rate of the global average. To help mitigate the impacts of accelerating sea ice loss, an Arctic sea ice forecasting AI system has been developed that outperforms traditional methods while running thousands of times faster.
The system, called 'IceNet', is based on a concept called deep learning, where the system 'learns' how sea ice changes from thousands of years of climate simulation data and decades of observational data. IceNet achieved state-of-the-art performance in seasonal forecasts of sea ice, a problem that has eluded scientists for decades. The initial study, published in the journal Nature Communications and funded by The Alan Turing Institute, resulted from a large collaboration of 17 researchers representing 11 different research centres across Europe and North America.
Work is now underway to develop the next generation of IceNet, which will run publicly in real-time, just like weather forecasts. This could be used as an early-warning system to help adapt to risks associated with rapid ice loss. James Byrne (BAS) and James Robinson (Alan Turing Institute) are developing the web application and cloud infrastructure to provide IceNet forecasts to end-users in the Arctic. Accurate sea ice forecasts could benefit a multitude of stakeholders. Indigenous communities could be informed when the sea ice might become dangerous to travel on. Regulatory bodies could define dynamic marine protected areas based on the predicted location of sea ice habitat.
Research vessels could better navigate around the edge of the ice pack, avoiding shipping disasters and saving lives. Conservation workers striving to protect vulnerable species like polar bears could better understand the challenges approaching in the coming weeks. The team at BAS is working with WWF to connect with these potential end-users to help realise the potential of IceNet as a smart conservation planning tool based on cutting-edge artificial intelligence.
Entrant: Tom Andersson , British Antarctic Survey (AI Lab)
Copyright: British Antarctic Survey
Links
https://eartharxiv.org/repository/view/2027/
https://www.turing.ac.uk/research/research-projects/understanding-arctic-sea-ice-loss