Artificial Intelligence Modeling Framework for Advancing Heliophysics Research (AIMFAHR)

Machine learning (ML) approaches are increasingly used in heliophysics to represent complex, coupled processes with greater computational efficiency than traditional physics-based models. As the number of data-driven models continues to grow, there is a need for frameworks that support their systematic integration and evaluation across multiple regions of the Sun–Earth system. The Artificial Intelligence Modeling Framework for Advancing Heliophysics Research (AIMFAHR) addresses this need by providing a modular, community-oriented environment for combining ML models into a unified geospace modeling capability.

Here we present initial efforts from the AIMFAHR model configuration and its application to storm-time geospace dynamics. The initial framework incorporates models spanning the magnetosheath, cusp regions, auroral precipitation, field-aligned currents (FACs), ionospheric electrodynamics, thermospheric density, and ground magnetic perturbations. Model behavior is examined during three geomagnetic storm events (4 January 2023, 6 May 2023, and 11 May 2024), selected as reference cases by the ML-based Geospace Environment Modeling (MLGEM) resource group at the Geospace Environment Modeling (GEM) workshop for the MLGEM Challenge Storm study.

AIMFAHR reproduces key features of storm-time responses across domains, including variations in dayside magnetic reconnection geometry and rates, cusp motion and ion energy dispersion, global auroral boundary evolution and spectral variability, enhanced FAC systems and ionospheric potentials, increased Joule heating and thermospheric density, and intensified ground magnetic disturbances. These results demonstrate the ability of an integrated, machine-learning-based framework to capture coherent system-level responses to solar wind driving. Ongoing AIMFAHR development focuses on enhancing the coupling between model components, transfer of learned representations across domains, quantification of predictive uncertainty, and transition toward operational space-weather applications.