Prediction of Geoeffective CMEs Using SOHO Images and Deep Learning

2024-11-01

Solar Physics

AI Summary

This study addresses the critical space weather problem of predicting the geomagnetic effectiveness of coronal mass ejections (CMEs). We propose the first end-to-end deep learning model that integrates multi-scale spatiotemporal convolution with Transformer-based self-attention mechanisms to directly learn discriminative features of CME-induced geomagnetic impacts from SOHO/LASCO C2/C3 image sequences—bypassing the limitations of handcrafted parameter extraction in conventional approaches. The architecture employs a ResNet-50 backbone augmented with 3D-CNNs to explicitly model temporal evolution, significantly enhancing early identification of intense geomagnetic storms. On the test set, the model achieves 89.2% accuracy and an AUC of 0.86, with warning lead times of 12–24 hours. It outperforms both state-of-the-art physics-based models and shallow machine learning methods, establishing a new paradigm for interpretable, high-temporal-resolution space weather forecasting.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Multi-modal Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

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

Problem

• Coronal Mass Ejections (CME)
• Geomagnetic Storms
• Space Weather Prediction

Innovation

• GeoCME
• Deep Learning
• Magnetic Storm Prediction