An artificial intelligence (AI) model that can predict changes in Arctic sea ice up to one year in advance has been developed, making it possible to provide mid- to long-term prediction information, which is expected to be helpful in developing Arctic shipping routes and exploring marine resources.

UNIST (President Jong-Rae Park) recently announced that Professor Jeong-Ho Lim's team from the Department of Earth Environmental and Urban Construction Engineering developed an artificial intelligence model that can predict the concentration of Arctic sea ice one year from now with an accuracy of less than 6%.

Sea ice concentration is the percentage of area covered by ice per unit area.

The research team developed this AI model by using UNET to learn the complex relationships between past patterns of Arctic sea ice concentration and key climate factors such as air temperature, water temperature, solar radiation, and wind.

UNET is one of the deep learning algorithms that AI uses to learn the relationship between image data such as satellite images.

The developed model had high mid- to long-term forecast accuracy. Accuracy was evaluated by comparing the predicted values of the AI model with past actual sea ice concentration values, and the average prediction error was less than 6% for all 3-month, 6-month, and 12-month predictions.

In the existing model, the average prediction error increased as the prediction period increased.

Additionally, the model showed stable prediction performance even in situations where sea ice decreased unusually rapidly.

In cases where sea ice melted rapidly, such as in the summers of 2007 and 2012, the existing model recorded an average prediction error of 17.35%, while the developed AI model recorded an average prediction error of 7.07%, reducing the average prediction error value by more than half.

The research team also identified climate factors that play an important role in mid- to long-term predictions of sea ice concentration.

Analysis of the differences between the UNET model prediction results showed that solar radiation and wind were the main variables at the sea ice edge where the ice thickness was thin.

Professor Lim Jeong-ho said, “This study has overcome the limitations of existing physics-based models and clarified the complex impact of various environmental factors on changes in Arctic sea ice,” adding, “It will be helpful in developing Arctic shipping routes, exploring marine resources, and establishing policies to respond to climate change.”

This study was published online on December 11 in the international journal Remote Sensing of Environment and was conducted with the support of the Korea Polar Research Institute and the Ministry of Oceans and Fisheries.