A team led by Niccolò Maffezzoli from Ca' Foscari University of Venice and the University of California, Irvine, has developed IceBoost, a global AI model for calculating glacier ice thickness distribution. The findings were published in Geoscientific Model Development. This model is expected to be a key tool for studying future glacier melt scenarios and predicting sea-level rise.
The IceBoost model combines decision tree algorithms trained on thickness measurements and 39 features, including ice velocity and temperature fields. According to Maffezzoli, the model demonstrates 30-40% lower errors than traditional models, especially in polar regions. The AI model leverages extensive observational data in conjunction with machine learning algorithms.
Accurate ice thickness estimates are crucial in polar regions and the margins of Greenland and Antarctica for modeling ice flow and projecting sea level rise. By the end of 2025, the researchers aim to release two datasets totaling half a million ice thickness maps, marking a significant step towards better understanding and predicting glacial impacts. This initiative aligns with the International Year of Glaciers' Preservation in 2025 and the Decade of Action for Cryospheric Sciences (2025 – 2034) declared by the United Nations.