Jacinda Knoll
Colorado Research Associates
Mentors: K. D. Leka and G. Barnes

 
Title: Statistical Prediction of Solar Flares Using Line of Sight Magnetogram Data


Abstract:

Solar flare prediction is becoming increasingly important as humans increase their presence in space and their dependence upon flare-sensitive activities on earth.  With the push for manned missions to the moon and Mars, the advent of space tourism, and the increasing numbers of polar flights, a reliable predictive measure for solar flares is sought by governments and private industry.

I analyze a large SOHO MDI dataset of line-of-sight magnetograms (prepared by J. McAteer) using the statistical technique of discriminant analysis.  I investigate dataset limitations and various corrective methods, as well as how the predictive power for solar flares varies with parameters chosen, year of data collection, and distance of the data from disk center.  

Use of line of sight magnetograms is limited because various corrective methods must be employed to compensate for lack of full vector information.  The two corrective models used in this study are a simple observing-angle correction and a potential field correction.  Parameters that perform best with this data set are the mean of the absolute value of the gradient of the magnetic field and variations on that parameter, as well as the total flux of the active region.  The angle-corrected data tend to perform better than the potential field correction, though this is ambiguous. 

When analyzed on an annual basis, certain years significantly outperform others.  It is unclear whether this trend is due to the number of magnetograms available from year to year or whether this outcome is related in some way to the solar cycle.  

Distance from disk center has a profound effect on predictive power, with skill scores falling nearly 200% between including data only within 5 degrees of disk center and including data through 45 degrees.  Further research is needed to analyze these findings in more detail.