Supervised convolutional neural networks for classification of flaring and nonflaring active regions using line-of-sight magnetograms

TL;DR

This study employs convolutional neural networks to classify solar active regions based on magnetogram data, revealing that flaring regions exhibit flare-productive states days before and after flares, with CNNs focusing on magnetic flux regions.

Contribution

The paper introduces a CNN-based approach for classifying flaring and nonflaring active regions using line-of-sight magnetograms, highlighting the importance of network design to avoid artifacts.

Findings

CNN achieves >60% recall in identifying flaring regions

CNN focuses on regions between opposite polarities and total unsigned flux

Magnetogram dimensions can introduce spurious dependencies

Abstract

Solar flares are explosions in the solar atmosphere that release intense bursts of short-wavelength radiation and are capable of producing severe space-weather consequences. Flares release free energy built up in coronal fields, which are rooted in active regions (ARs) on the photosphere, via magnetic reconnection. The exact processes that lead to reconnection are not fully known and therefore reliable forecasting of flares is challenging. Recently, photospheric magnetic-field data has been extensively analysed using machine learning (ML) and these studies suggest that flare-forecasting accuracy does not strongly depend on how long in advance flares are predicted (Bobra & Couvidat 2015; Raboonik et al. 2017; Huang et al. 2018). Here, we use ML to understand the evolution of AR magnetic fields before and after flares. We explicitly train convolutional neural networks (CNNs) to classify SDO/HMI line-of-sight magnetograms into ARs producing at least one M- or X-class flare or as nonflaring. We find that flaring ARs remain in flare-productive states -- marked by recall >60% with a peak of ~ 80% -- days before and after flares. We use occlusion maps and statistical analysis to show that the CNN pays attention to regions between the opposite polarities from ARs and the CNN output is dominantly decided by the total unsigned line-of-sight flux of ARs. Using synthetic bipole magnetograms, we find spurious dependencies of the CNN output on magnetogram dimensions for a given bipole size. Our results suggest that it is important to use CNN designs that eliminate such artifacts in CNN applications for processing magnetograms and, in general, solar image data.