Researchers at Sun Yat-sen University, in collaboration with the South Guangdong (Zhuhai) Marine Science and Engineering Laboratory in China, have made an accurate prediction regarding Antarctic sea ice from December 2023 to February 2024 using deep learning methods.
Glacier in Antarctica. (Photo (archive): AFP/TTXVN)
The research team utilized a Long Short-Term Memory Convolutional Neural Network (ConvLSTM) to construct a seasonal model for predicting Antarctic sea ice. The model indicates that Antarctic sea ice levels will remain close to historical lows in February 2024, with few signs of reaching a new record low. The predicted sea ice area (SIA) and sea ice extent (SIE) for February 2024 are 1.441 million km² and 2.105 million km², respectively, slightly higher than the historical lows recorded in 2023.
The team made their predictions in December 2023, and this information was published in the journal Advances in Atmospheric Sciences in February of this year. Their predictions were subsequently confirmed by the latest satellite observations in the same month, with SIA and SIE measuring 1.510 million km² and 2.142 million km², respectively.
According to the researchers, the satellite observations showed minimal deviation from the predicted values, indicating reliability in forecasting Antarctic sea ice and the accuracy of the ConvLSTM model.
Professor Yang Qinghua at Sun Yat-sen University emphasized that the successful prediction by the research team not only highlights the importance of enhancing sea ice forecasting research in Antarctica but also demonstrates the significant application potential of deep learning methods in this crucial field.
