Solar Flare Detection from Sudden Ionospheric Disturbances in VLF Signals via a CNN-HMM Framework

Area of Science:

• Space Physics
• Artificial Intelligence
• Signal Processing

Background:

• Solar flares pose risks to space and terrestrial technologies.
• Current monitoring relies heavily on satellite-based systems.
• Sudden Ionospheric Disturbances (SIDs) are reliable indicators of solar flare activity.

Purpose of the Study:

• To develop an automated system for detecting solar flares (>=M1.0) using Very Low Frequency (VLF) signals.
• To leverage hybrid AI models for improved accuracy and temporal consistency in flare detection.
• To explore a low-cost, ground-based alternative or complement to satellite monitoring.

Main Methods:

• A hybrid Convolutional Neural Network-Hidden Markov Model (CNN-HMM) framework was developed.
• CNN processed VLF signal windows and derivatives for flare probability estimates.
• HMM with Viterbi decoding ensured temporal consistency and physically plausible event sequences.

Main Results:

• The model achieved a balanced accuracy of 0.819 and MCC of 0.529 at the sample level.
• Event-level detection reached a peak F1-score of 0.558 for flares >=C6.0.
• Demonstrated automated, physically consistent detection of solar flares via SIDs.

Conclusions:

• The CNN-HMM framework provides effective automated detection of solar flares using VLF signals.
• The approach shows potential for real-time early warning systems.
• Combining data from multiple receivers could create a cost-effective monitoring network.