AI reconstructs Earth's mantle flow history with minimal abundance relevance

University of Tsukuba researchers used AI to reconstruct Earth's mantle flow history from surface data, but the study explicitly states this approach offers minimal direct benefits for practical abund

University of Tsukuba researchers developed an AI model using physics-informed neural networks to reconstruct Earth's mantle flow history from surface observations. The model was trained on synthetic near-surface data and present-day temperature measurements, then validated against thermal convection simulations. Published in the Journal of Geophysical Research: Machine Learning and Computation (2026), the study confirms high accuracy in reconstructing historical mantle convection features but explicitly states minimal direct abundance benefits for practical applications. Crucially, the model was trained on synthetic data rather than real-world geophysical observations, and applying it to actual Earth data requires further development. This research enhances geological hazard understanding but does not currently impact the availability of basic needs like clean water, energy, or shelter through direct mechanisms. The source provides no detail on how this work might eventually scale toward abundance goals.

Source: Phys.org