A team of scientists from Norway and the United Kingdom has trained an artificial intelligence (AI) system to accurately map the surface area and outlines of giant icebergs in satellite images in hundredths of a second. The study was published in the scientific journal magazine Cryosphere.
Manual processing of satellite images by a person is more accurate but takes up to a few minutes. This needs to be repeated many times, making the process labor-intensive and time-consuming; The neural network copes with this task 10 thousand times faster than a human.
The team used an algorithm called U-net to “train” a computer to accurately map the outlines of icebergs based on images taken by the European Space Agency’s Sentinel-1 satellites.
The system was tested on satellite images of seven icebergs ranging in size from 54 square kilometers (the area of the city of Bern in Switzerland) to 1052 square kilometers (comparable to the area of Hong Kong in China).
Up to 46 images were used for each of these icebergs, covering all seasons 2014-2020.
In a series of tests, U-net outperformed the other two algorithms and was more effective at identifying the outline of an iceberg in images taken in harsh environmental conditions, such as an image capturing many ice structures.
Scientists say the technology could lead to new services that provide information about the shape and size of giant icebergs.
Icebergs have a significant impact on the polar environment, and monitoring them is critical for both maritime safety and scientific research. They can be extremely large (in some cases the size of a small country) and pose a danger to passing ships. As icebergs melt, they release nutrients and freshwater into the seas, which can impact marine ecosystems.
Previous scientists warned About the irreversible melting of the West Antarctic ice sheet.