Abstract

A test of a new Bayesian approach to solar flare prediction is presented. The approach uses the past history of flaring together with phenomenological rules of flare statistics to make a prediction for the probability of occurrence of a large flare within an interval of time, or to refine an initial prediction (which may incorporate other information). The test of the method is based on data from the Geostationary Observational Environmental Satellites, and involves whole-Sun prediction of soft X-ray flares for 1976–2003. The results show that the method somewhat over-predicts the probability of all events above a moderate size, but performs well in predicting large events.