Onset-centered complexity signatures of GOES flare class in Konus-WIND G1–G3 time series

Abstract

The high-energy emission from solar flares demonstrates highly variable and often complex light curves, whose temporal structure may contain information beyond standard amplitude-based summaries. Using the Konus-WIND solar flare observations, we carried out an investigation into the statistical association between the onset-centered changes in geometric and complexity-based descriptors and GOES flare classes. Our study was based on the data from the background-subtracted count rates from the event catalogue in three energy channels: G1 (20--80~keV), G2 (80--300~keV) and G3 (300--1200~keV). The observation of each event was segmented into pre-onset (``before'') and post-onset (``after'') intervals defined by bin boundaries relative to the trigger time. For each channel and segment we computed five descriptors: Higuchi fractal dimension, normalized permutation entropy, statistical complexity \(C_{JS}\) in the complexity--entropy framework, normalized Lempel--Ziv complexity (median-binarized), and Hjorth complexity. We summarized onset-associated changes by paired differences \(\Delta=\mathrm{after}-\mathrm{before}\) and assessed class dependence using nonparametric hypothesis tests with Benjamini--Hochberg false-discovery-rate control.

In different channels, most descriptors showed non-zero onset-associated shifts, and the magnitude of \(\Delta\) was significantly different for most channel--descriptor combinations across GOES classes. The strongest monotonic class scaling was observed in the G1 channel for Higuchi fractal dimension, permutation entropy, and \(C_{JS}\), while the clearest separation in the G2 channel was provided by normalized Lempel--Ziv and Hjorth complexity, including a marked Hjorth-complexity transition for X-class events. In the G3 channel, onset-related changes remained evident but between-class discrimination was weaker overall, with Higuchi fractal dimension showing near class-invariant $\Delta$. The obtained results provide conservative population-level evidence that multi-channel onset-centered complexity and geometric descriptors are statistically related to flare class, and motivate future work on integrating additional physical covariates and strictly validated classification models.