SOURCE-LINKED INTELLIGENCE

SOFIA: Intelligent analysis of solar activity for space weather forecasting

CORDIS · observation · Publication date unknown

ng systems and safeguard European infrastructure, yet current approaches lack consistent parameters that capture how solar activity emerges and evolves across scales. This project integrates advanced machine learning methods to improve the description and prediction of solar activity. A central innovation is the extensive use of Variational Autoencoders (VAEs) for explainable ML-based parametrization of complex solar processes. First, VAEs will generate structured latent representations of solar active regions from multi-wavelength observations provided by NASA’s Solar Dynamics Observatory, delivering physically meaningful parameters that link magnetic morphology to the occurrence of eruptions. Weakly supervised and sequential models will then enhance interpretability and capture temporal evolution, while extensions to full-disk solar observations will integrate local and global drivers of solar variability. Finally, these data-driven representations will be coupled with geophysical models of the near-Earth environment, connecting solar drivers to response in interplanetary space and

Read original source ↗ Open in workspace

recordType

award

status

SIGNED

region

EU

value

202125.12

unit

EUR