Answers are grounded in this paper and cite their sources.
Solar flares are explosive coronal events that influence near-Earth space weather. Their prediction remains challenging because photospheric magnetic conditions governing flare onset, including current helicity, remain incompletely characterized. Empirical models and deep learning approaches based on magnetic parameters have advanced solar flare forecasting, yet the former rely on simplified statistics and the latter act as black boxes, leaving the physical origin of their predictions unclear. In this letter, we present interpretable major (≥M - class) flare forecasting models that combine deep learning with symbolic regression (SR) using time series active region (AR) magnetic parameters. Our SR models achieve true skill statistics (TSS) and a Heidke skill score (HSS) of 0.50 ± 0.17 and 0.36 ± 0.04, fully comparable to deep learning baselines (TSS 0.38–0.56 and HSS 0.30–0.54), using analytical formulae with over 4 orders of magnitude fewer degrees of freedom. A representative formula reveals that flare probability is driven by magnetic flux near the polarity inversion line ( R -value) through a nonlinear gating mechanism, activated only when the product of total and absolute current helicity exceeds a critical threshold. This helicity interaction term, which is strongly correlated with the squared sum of current helicity at the largest spatial scales, accounts for ∼77% of the predicted variance. This formula successfully identifies many major flares in ARs 12673 and 13664 missed by the Community Coordinated Modeling Center flare scoreboard models. These results suggest that large-scale helicity accumulation serves as a primary physical driver of major flare onset. This finding has immediate implications for operational space weather forecasting.
No takes yet. Share an insight, caveat, or question.
Kim et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: