Scientists have developed an artificial intelligence (AI) system capable of accurately mapping the surface area and outline of giant icebergs captured on satellite images within one-hundredth of a second. This breakthrough, outlined in a paper titled “Mapping the extent of giant Antarctic icebergs with Deep Learning,” addresses the limitations of existing automated systems that struggle to distinguish icebergs from other features in the image.

Manual interpretation of the images can provide more accurate results, but it is a time-consuming and laborious process. It can take several minutes to delineate the outline of a single iceberg. Therefore, the new AI system greatly streamlines and accelerates the mapping process.

Monitoring icebergs is crucial for maritime safety and scientific research, as they have a significant impact on the polar environment. They can reach vast sizes, comparable to small countries, and pose risks to passing ships. Additionally, as icebergs melt, they release freshwater and nutrients into the seas, which can affect marine ecosystems.

The AI system, developed by Dr. Anne Braakmann-Folgmann and her team, utilizes an algorithm called U-net, which is a type of neural network. The researchers trained the computer using satellite images taken by the European Space Agency’s Sentinel-1 satellites.

In comparative tests, the U-net algorithm outperformed two other state-of-the-art algorithms, namely k-means and Otsu, which are commonly used for iceberg mapping. The U-net algorithm proved effective at distinguishing icebergs from other ice floating on the sea or surrounding coastlines, even in complex images with overlapping features.

The successful implementation of this AI system provides researchers with a valuable tool to observe and understand the behavior of icebergs, including the melting and fragmentation processes. With the ability to rapidly and accurately map icebergs, scientists can gain valuable insights into the impact of these massive structures on the Earth’s polar regions.