An automated classification approach to ranking photospheric proxies of magnetic energy build-up

¹ Klipsch School of Electrical & Computer Engineering, New Mexico State University, Las Cruces, NM 88003, USA
e-mail: amani@nmsu.edu; lboucher@nmsu.edu
² Department of Astronomy, New Mexico State University, Las Cruces, NM 88003, USA
e-mail: mcateer@nmsu.edu

Received: 26 February 2015
Accepted: 13 May 2015

Abstract

Aims. We study the photospheric magnetic field of ~2000 active regions over solar cycle 23 to search for parameters that may be indicative of energy build-up and its subsequent release as a solar flare in the corona.

Methods. We extract three sets of parameters: (1) snapshots in space and time: total flux, magnetic gradients, and neutral lines; (2) evolution in time: flux evolution; and (3) structures at multiple size scales: wavelet analysis. This work combines standard pattern recognition and classification techniques via a relevance vector machine to determine (i.e., classify) whether a region is expected to flare (≥C1.0 according to GOES). We consider classification performance using all 38 extracted features and several feature subsets. Classification performance is quantified using both the true positive rate (the proportion of flares correctly predicted) and the true negative rate (the proportion of non-flares correctly classified). Additionally, we compute the true skill score which provides an equal weighting to true positive rate and true negative rate and the Heidke skill score to allow comparison to other flare forecasting work.

Results. We obtain a true skill score of ~0.5 for any predictive time window in the range 2 to 24 h, with a true positive rate of ~0.8 and a true negative rate of ~0.7. These values do not appear to depend on the predictive time window, although the Heidke skill score (<0.5) does. Features relating to snapshots of the distribution of magnetic gradients show the best predictive ability over all predictive time windows. Other gradient-related features and the instantaneous power at various wavelet scales also feature in the top five (of 38) ranked features in predictive power. It has always been clear that while the photospheric magnetic field governs the coronal non-potentiality (and hence likelihood of producing a solar flare), photospheric magnetic field information alone is not sufficient to determine this in a unique manner. Furthermore we are only measuring proxies of the magnetic energy build up. We are still lacking observational details on why energy is released at any particular point in time. We may have discovered the natural limit of the accuracy of flare predictions from these large scale studies.