IceBoost AI model for global glacier mapping

Researchers from Italy's Ca' Foscari University of Venice and the National Research Council's Institute for Polar Sciences, in collaboration with international partners including NASA's Jet Propulsion Laboratory, have developed IceBoost v2.0. This machine-learning model is designed to map global terrestrial glacier ice volumes by analyzing over seven million thickness measurements and integrating 26 physical and geometric variables, such as terrain slope, curvature, ice flow velocity, and local temperature.

Published in Scientific Data (Nature), the model improves volume estimate accuracy for the global glacier inventory by up to 40% compared to previous methods, though it excludes the Antarctic and Greenland ice sheets. The study estimates global glacier ice at approximately 150,000 cubic kilometers, which would lead to a global mean sea-level rise of 32.3 cm if entirely melted.

The researchers have released an interactive web application to allow experts and the public to explore the high-resolution mapping data. This information is intended to support future glacier simulations and IPCC assessments, particularly regarding freshwater availability for approximately 1.9 billion people.