Artificial intelligence helps scientists better predict glacier evolution in response to climate change

Glacier mass is currently being lost due to human-induced climate change. It is extremely important to understand the physical processes related to these regional and global changes in order to anticipate possible future glacier developments and their impacts on sea level rise, water resources and ecosystems. To address these issues, numerical models allow scientists to simulate, in a simplified way, the evolution of glaciers for entire regions or over the entire planet, both for past and future periods.

In a new study published in the journal Nature Communications, an interdisciplinary team of glaciologists, climatologists and mathematicians from the University of Grenoble Alpes, INRAE, Météo-France, the Free University of Brussels and TU Delft used for the first time deep artificial neural networks - a type of artificial intelligence - to simulate the future evolution of glaciers on a regional scale. Like most physical processes in nature, the evolution of glaciers in response to climate is non-linear, i.e. it does not evolve consistently over time. The ability to capture these non-linear behaviours is precisely one of the advantages that neural networks offer over the models currently used for simulations on regional to global scales. This study thus marks the beginning of a new generation of scientific models that are more powerful and better suited to predicting the future evolution of glaciers in the face of climate change.

These results have important implications for our knowledge of future glacier and sea level change. They predict that glaciers in the Arctic and Patagonia, which contain the world’s largest ice reserves outside of Greenland and Antarctica, would be most affected by this non-linear response to warming, calling for a revision of current predictions, with models capable of better reproducing the non-linearities of their future evolution. The use of artificial intelligence, combined with climate and glacier physics, will play a key role in future findings.

Authors : Jordi Bolibar, Antoine Rabatel, Isabelle Gouttevin, Harry Zekollari, Clovis Galiez - 20/01/2022
Image : Artistic representation of the artificial intelligence model, based on a deep neural network, used to model the future evolution of glaciers (in this image : Aoraki/Mount Cook, New Zealand). The individual nodes represent the artificial neurons, and the numbers represent the data used to train the model. Attribution : photo : Tom Bernardo, artistic representation : Jordi Bolibar

Link to article : https://www.nature.com/articles/s41467-022-28033-0

Read articles published by UGA and INSU-CNRS.

Translated with www.DeepL.com/Translator (free version)