Speaker
Description
We develop a Convolutional Neural Network (CNN)-based framework to improve solar flare identification from soft X-ray (SXR) observations and investigate flare size-waiting-time correlations, where by size we mean SXR peak flux. We study the statistical properties of solar flares both globally (i.e., across the solar disk) and locally (i.e., within individual active regions, ARs) and compare the results with previous studies. Our aim is to place the observations in the context of competing models of flare triggering and magnetic energy release. The CNN-based framework has established the most comprehensive record of solar flares to date, containing more than seven times as many events as existing catalogs. To associate the CNN-detected SXR events with their ARs of origin, we develop a probabilistic framework and use Solar Dynamics Observatory extreme ultraviolet images. A Bayesian blocks analysis of the waiting-time distributions indicates broad consistency with a piecewise Poisson process. Size-waiting-time correlations are evaluated using time symmetry tests (i.e., tests of the statistical equivalence of waiting times before and after a flare), the Spearman rank correlation, and the Efron-Petrosian test for truncated data. No statistically significant correlations between flare sizes and waiting times are detected in most studied ARs at most times, although specific instances of correlation are identified in some regions. We show that previously reported correlations between flare sizes and waiting times are significantly influenced by obscuration, that is, under-counting weaker or overlapping flares during periods of elevated flux.
| I am the presenting author | No |
|---|---|
| If you are not the presenting author, please give the presenting author's name: | Michael Wheatland |