Chinese scientists accurately predict sea ice in Antarctica thanks to machine learning

Researchers at Sun Yat-sen University in conjunction with China's Guangdong (Zhuhai) Southern Marine Science and Engineering Laboratory have made accurate predictions about Antarctic sea ice since December. 2023- February 2024 thanks to deep machine learning methods.

Glacier in Antarctica. (Photo (document): AFP/TTXVN)

The research team used a long-term oriented short-term neural network (ConvLSTM) to build a seasonal-scale Antarctic sea ice prediction model. The model indicates that Antarctic sea ice will remain near historic lows in February 2024, but there are few signs of reaching new record lows. The predicted sea ice area (SIA) and sea ice extent (SIE) in February 2024 are 1.441 million km2 and 2.105 million km2, respectively, slightly higher than the historical low recorded in 2023.

The research team made the above prediction in December 2023 and this information was published in the journal Advances in Atmospheric Sciences in February this year. Their prediction was later confirmed by the latest satellite observations in the same month, with SIA and SIE areas of 1.510 million square kilometers and 2.142 million square kilometers, respectively.

The research team used a long-term oriented short-term neural network (ConvLSTM) to build a seasonal-scale Antarctic sea ice prediction model. The model indicates that Antarctic sea ice will remain near historic lows in February 2024, but there are few signs of reaching new record lows. The predicted sea ice area (SIA) and sea ice extent (SIE) in February 2024 are 1.441 million km ² and 2.105 million km ² , respectively, slightly higher than the historical low recorded in 2023. .

The research team made the above prediction in December 2023 and this information was published in the journal Advances in Atmospheric Sciences in February this year. Their prediction was later confirmed by the latest satellite observations in the same month, with SIA and SIE areas of 1.510 million km ² and 2.142 million km ² respectively .

According to the researchers, the results recorded through satellite observations are not much different from the predicted values , which shows the reliability of Antarctic sea ice forecasts and the accuracy of the ConvLSTM model. .

Professor Yang Qinghua at Sun Yat-sen University emphasized that the team's successful prediction not only emphasizes the importance of strengthening sea ice prediction research in Antarctica but also demonstrates significant application potential. of deep machine learning methods in this important field.