Predicting Associations between Solar Flares and Coronal Mass Ejections Using SDO/HMI Magnetograms and a Hybrid Neural Network

2026-02-19

Astrophysical Journal

AI Summary

Predicting whether a solar flare will be accompanied by a coronal mass ejection (CME) remains challenging due to an incomplete understanding of their underlying physical linkage. This work proposes a hybrid neural network that integrates Vision Transformers with Long Short-Term Memory (LSTM) networks to analyze time-series line-of-sight magnetograms from SDO/HMI, capturing spatiotemporal evolution patterns in active regions to forecast whether an upcoming flare within the next 24 hours will be eruptive (associated with a CME) or confined (without a CME). To our knowledge, this is the first application of a Vision Transformer–LSTM architecture to flare–CME association prediction, demonstrating superior performance on this task. The model further highlights the critical role of magnetic flux cancellation near polarity inversion lines in triggering CME-associated flares, offering new observational evidence for the physical mechanisms driving eruptive flares.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Solar eruptions, including flares and coronal mass ejections (CMEs), have a significant impact on Earth. Some flares are associated with CMEs, and some flares are not. The association between flares and CMEs is not always obvious. In this study, we propose a new deep learning method, specifically a hybrid neural network (HNN) that combines a vision transformer with long short-term memory, to predict associations between flares and CMEs. HNN finds spatio-temporal patterns in the time series of line-of-sight magnetograms of solar active regions collected by the Helioseismic and Magnetic Imager on board the Solar Dynamics Observatory and uses the patterns to predict whether a flare projected to occur within the next 24 hr will be eruptive (i.e., CME-associated) or confined (i.e., not CME-associated). Our experimental results demonstrate the good performance of the HNN method. Furthermore, the results show that magnetic flux cancellation in polarity inversion line regions may well play a role in triggering flare-associated CMEs, a finding consistent with the literature.

Problem

• solar flares
• coronal mass ejections
• flare-CME association
• eruptive flares
• confined flares

Innovation

• hybrid neural network
• vision transformer
• long short-term memory
• solar flare–CME association
• SDO/HMI magnetograms