For the first time, researchers have used machine learning to analyze data on the movement of ice in Antarctica, obtained remotely using satellite imagery and radar from aircraft. The analysis revealed that 95% of the ice shelf is anisotropic. This means that its physical properties vary in different directions.
Movement of glaciers in Antarctica
Existing models inaccurately reflect the movement of the ice sheet, which is observed from satellites. The identification of patterns governing the movement of the Antarctic ice sheet will help in the future to more accurately predict the changes taking place on the continent. The results of the work are published in the journal Science.
Antarctica is a continent almost completely covered with ice. Antarctica’s glaciers contain about 90% of the world’s freshwater reserves. The area of the ice sheet is about 14 million square kilometers (like two areas of Australia). The thickness of the ice can reach 4.7 km. Due to global warming, Antarctic ice is melting, leading to sea level rise worldwide.
Previously, researchers used models based on laboratory experiments to study the movement of ice in Antarctica, but they did not take into account all the features of Antarctic ice. Ice formed from seawater has different properties than ice formed from compacted snow. Real ice sheets may contain large cracks, air pockets, or other inconsistencies that affect movement.
Further reduction of the ice sheet may slow down important ocean currents, which will affect global weather patterns and ocean temperatures. Scientists used machine learning to analyze the movement of ice in Antarctica from 2007 to 2018. The data was obtained remotely using satellite imagery and radar from aircraft. This made it possible to study the movements and thickness of the ice on a large scale.
The machine learning model took into account the laws of physics that control the movement of ice. Thanks to this, scientists were able to obtain new data on the viscosity of ice, that is, on its resistance to movement. The researchers focused on five Antarctic ice shelves. Ice shelves are floating platforms of ice that extend into the ocean from land glaciers and hold back most of Antarctica’s glacial ice.
Scientists have discovered that parts of glaciers located closer to the continent are shrinking. This is consistent with the results of laboratory experiments where ice samples were compressed to determine their strength characteristics. However, the further away the ice is from the continent, the more it stretches towards the sea. At the same time, its physical properties become different in different directions, and this phenomenon is called anisotropy.
Scientists have found that most of the ice shelf is anisotropic. The compression zone occupies only a small part (less than 5%), and the rest (95%) is located in the stretch zone and obeys other laws. Most models assume that Antarctic ice has the same properties in all directions. Researchers have shown that this is not the case, and existing models inaccurately reflect the movement of ice observed from satellites.
The melting of glaciers leads to a redistribution of the Earth’s mass, slowing down its rotation and shifting the axis of rotation. In addition, an increase in the amount of meltwater can lead to instability or destruction of the ice shelf, which, in turn, will lead to sea level rise.”We are trying to show that with the help of artificial intelligence, you can learn something new.
It still needs to be bound by some physical laws, but this combined approach has allowed us to discover the physics of ice beyond what was previously known, and can really contribute to a new understanding of terrestrial and planetary processes in natural conditions,” said Ching-Yao Lai from Stanford University.
By Oksana Gritsenko
