Authors: Jialiang Li (Institute for Space Weather Sciences, New Jersey Institute of Technology), Vasyl Yurchyshyn (Big Bear Solar Observatory, New Jersey Institute of Technology), Jason T. L. Wang (Institute for Space Weather Sciences, New Jersey Institute of Technology), Haimin Wang (Institute for Space Weather Sciences, New Jersey Institute of Technology), Hongyang Zhang (Institute for Space Weather Sciences, New Jersey Institute of Technology)

Coronal mass ejections (CMEs), also known as solar storms, propel the mass and magnetic
substances from the Sun into interplanetary space on a rapid timescale. These eruptive events
emanate from solar active regions (ARs), which are characterized by pronounced magnetic
fields that exhibit dynamical changes. CMEs can disrupt crucial technologies on Earth or in
the near-Earth space environment. Among the CMEs, halo CMEs, including full-halo CMEs
and partial-halo CMEs, are of particular importance due to their Earth-directed propagation
tendencies. Here, we propose advanced artificial intelligence (AI) techniques, including novel
generative AI (GenAI) and explainable AI (XAI) methods, to predict CMEs. First, we present
a GenAI tool to enlarge the sample size by temporally super-resolving SOHO/MDI
magnetograms of solar ARs from a cadence of 96 minutes to 12 minutes at the level of
SDO/HMI magnetograms. Then, we describe an XAI method to predict associations between
solar flares and CMEs, classifying eruptive and confined flares, using SDO/HMI
magnetograms. Finally, we present an XAI tool to predict halo CMEs and estimate their
geoeffectiveness using SDO/HMI vector magnetic data products. These XAI methods can be
extended to an augmented harmonized dataset that includes SDO/HMI magnetograms and
spatially/temporally super-resolved SOHO/MDI magnetograms of solar ARs to enhance
prediction accuracy. Extensive experiments demonstrate the good performance and
usefulness of the proposed AI methods.