Predicting Solar Flares Using CNN and LSTM on Two Solar Cycles of Active Region Data

TL;DR

This study employs CNN and LSTM deep learning models to predict solar flares using data from two solar cycles, demonstrating improved accuracy with ensemble methods and interpretability via attribution techniques.

Contribution

It introduces a stacking ensemble of CNN and LSTM for flare prediction and applies Integrated Gradients for model interpretability, highlighting the importance of multi-cycle data and magnetic flux features.

Findings

LSTM trained on two cycles outperforms single-cycle training with higher TSS.

Ensemble of CNN and LSTM yields higher TSS than individual models.

Integrated Gradients attribute flare predictions to magnetic flux, revealing model limitations.

Abstract

We consider the flare prediction problem that distinguishes flare-imminent active regions that produce an M- or X-class flare in the future 24 hours, from quiet active regions that do not produce any flare within $\pm 24$ hours. Using line-of-sight magnetograms and parameters of active regions in two data products covering Solar Cycle 23 and 24, we train and evaluate two deep learning algorithms -- CNN and LSTM -- and their stacking ensembles. The decisions of CNN are explained using visual attribution methods. We have the following three main findings. (1) LSTM trained on data from two solar cycles achieves significantly higher True Skill Scores (TSS) than that trained on data from a single solar cycle with a confidence level of at least 0.95. (2) On data from Solar Cycle 23, a stacking ensemble that combines predictions from LSTM and CNN using the TSS criterion achieves significantly higher TSS than the "select-best" strategy with a confidence level of at least 0.95. (3) A visual attribution method called Integrated Gradients is able to attribute the CNN's predictions of flares to the emerging magnetic flux in the active region. It also reveals a limitation of CNN as a flare prediction method using line-of-sight magnetograms: it treats the polarity artifact of line-of-sight magnetograms as positive evidence of flares.