False Alarms in the Forecasting of Solar Energetic Particle Events

Abstract

Solar Energetic Particles (SEPs) are known to be accelerated by high-energy events in the Sun's corona: coronal mass ejections (CMEs) with high speed, solar flares with high peak emission in soft X-rays, or a combination of the two. SEPs, however, are not detected following all fast CMEs or intense flares. Those large solar events, which might reasonably have been expected to produce SEPs at Earth but which failed to do so, may be termed “false alarms”.
In this work, two simple SEP forecasting algorithms are defined: one (algorithm A.1) is based upon the observation of a magnetically well-connected CME with a speed of 1,500 km/s or greater (a ``fast CME''), and the other (algorithm A.2) is based upon the observation of a magnetically well-connected X class flare. The algorithms were applied to historical data sets to ascertain which produced an enhancement of >40 MeV protons, and which were false alarms.
The algorithms have been evaluated using standard verification scores. Both algorithms correctly forecast approximately the same percentage of SEP events (47% and 49% respectively); the false alarm ratio for algorithm A.1, however, was much lower than for A.2 (29% and 51% respectively). Both algorithms failed to forecast almost the same number of SEP events (53% for A.1, and 51% for A.2).

The parameters of the false alarms were compared to those of the SEP-producing events. False alarm fast CMEs tended to be associated with flares of class less than M3; X class flares which were either not associated with any CME, or were associated with a CME slower than 500 km/s, were false alarms.
A third forecasting algorithm, based upon these results, was defined. This algorithm, which takes into account parameters of both CMEs and flares, performed better than either A.1 or A.2, correctly forecasting a significantly greater percentage of SEP events than both (68%), having a false alarm ratio similar to A.1 (30%), but missing a significantly lower percentage (32%) of SEP events.
A small number of case studies were carried out. It was found that for accurate forecasting of SEP events it may not be sufficient simply to consider the accelerating events, but that the location of the heliospheric current sheet relative to the site of the solar event and of the Earth's footpoint may be an important consideration.
SEP forecasts produced by the SPARX simulation were evaluated with a view to providing a benchmark against which future versions of the model may be tested.