Photospheric Magnetic Field Properties of Flaring vs. Flare-Quiet Active Regions I: Data, General Approach, and Statistical Results
Abstract
Photospheric vector magnetic field data from the U. Hawai`i Imaging Vector Magnetograph are examined for pre-event signatures unique to solar energetic phenomena. Parameters are constructed from B(x,y) to describe (for example) the distributions of the field, spatial gradients of the field, vertical current, current helicity, ''twist'' parameter α and magnetic shear angles. A quantitative statistical approach employing discriminant analysis and Hotelling's T2-test is applied to the magnitude and temporal evolution of parameters from 24 flare-event and flare-quiet epochs from seven active regions.
We demonstrate that (1) when requiring a flare-unique signature, numerous candidate parameters are nullified by considering flare-quiet epochs, (2) a more robust method exists for estimating error rates than conventional ''truth tables'', (3) flaring and flare-quiet populations do not necessarily have low error rates for classification even when statistically distinguishable, and that (4) simultaneous consideration of a large number of variables is required to produce acceptable error rates. That is, when the parameters are considered individually, they show little ability to differentiate between the two populations; multi-variable combinations can discriminate the populations and/or result in perfect classification tables. In lieu of constructing a single all-variable discriminant function to quantify the flare-predictive power of the parameters considered, we devise a method whereby all permutations of the four-variable discriminant functions are ranked by Hotelling's T2. We present those parameters (e.g. the temporal increase of the kurtosis of the spatial distribution of the vertical current density) which consistently appear in the best combinations, indicating that they may play an important role in defining a pre-event photospheric state. While no single combination is clearly the best discriminator, we demonstrate here the requisite approach: include flare-quiet epochs as a control group for statistical tests of the null hypothesis. This work was performed under Air Force Office of Scientific Research contracts F49620-00-C-0004 and F49620-03-C-0019.- Publication:
-
AAS/Solar Physics Division Meeting #34
- Pub Date:
- May 2003
- Bibcode:
- 2003SPD....34.1615L