Toward Reliable Benchmarking of Solar Flare Forecasting Methods 

D. Shaun Bloomfield Submitted: 20120227 09:04
Solar flares occur in complex sunspot groups, but it remains unclearhow the probability of producing a flare of a given magnitude relatesto the characteristics of the sunspot group. Here, we useGeostationary Operational Environmental Satellite Xray flares andMcIntosh group classifications from solar cycles 21 and 22 tocalculate average flare rates for each McIntosh class and use these todetermine Poisson probabilities for different flare magnitudes.Forecast verification measures are studied to find optimum thresholdsto convert Poisson flare probabilities into yes/no predictions ofcycle 23 flares. A case is presented to adopt the true skill statistic(TSS) as a standard for forecast comparison over the commonly usedHeidke skill score (HSS). In predicting flares over 24 hr, the maximumvalues of TSS achieved are 0.44 (Cclass), 0.53 (Mclass), 0.74(Xclass), 0.54 (>=M1.0), and 0.46 (>=C1.0). The maximum values of HSSare 0.38 (Cclass), 0.27 (Mclass), 0.14 (Xclass), 0.28 (>=M1.0), and0.41 (>=C1.0). These show that Poisson probabilities performcomparably to some more complex prediction systems, but the overallinaccuracy highlights the problem with using average values torepresent flaring rate distributions.
Authors: D. Shaun Bloomfield, Paul A. Higgins, R. T. James McAteer, and Peter T. Gallagher
Projects: None

Publication Status: Published in ApJL
Last Modified: 20120229 08:25


