Flare & group evolution prediction tables

Attached tables provide an overview on the likelihood that a sunspotgroup will generate a flare, and the likelihood it will evolve into another McIntosh-classification. These results are based on data from the 23^rd solar cycle (1996-2006).

McIntosh-classifications (McI) per day and per sunspotgroup were used from the USAF network. A description of McI was published in Solar Physics (1990). It is implicitely assumed that the classification made by the USAF stations is always correct.

The flare-data were taken from the NOAA/NGDC database (GOES). Only flares that were linked to a particular sunspotgroup were used.

Table 1 provides an overview of the data from the USAF network stations. From the total of over 63000 classifications, selections were made in such a way that there was always 1 McI for every day that the sunspotgroup was visible. Stations were chosen in order of highest yearly number of observations. This makes Learmonth the prime contributor, supported with data from Ramey AFB and Holloman AFB.

3 McI were added in order to cover the occurence of flares when the concerned sunspot was not visible:

• XXX: an event occured before the sunspotgroup became visible for the first time, for example when it was still behind the east limb.
• YYY: an event occured during a period when the sunspot was not visible, for example a small sunspotgroup which is not visible for some days then to reappear for one day. This classification covers also most days when no data were available from the network (exceptionnal).
• ZZZ: an event occured after the sunspot was last visible, for example, when it had already rounded the west limb.

Some corrections were made to the McI-data (typo's), and doubles (multiple reports from the same station for one group in one day) were eliminated, taking care that the classification was in line with previous and/or following days. In total, over 22000 McI were used for flare data, and more than 19000 for the transitions.

Results were expressed in % likelihood that a flare would be produced in the next 24 hours, or that a sunspot group would evolve from one McI into another. The algorithm used is the same as that on the Solar Monitor webpage, which in turn is based on methodologies described by Wheatland and Moon et al.. Other methodologies and discussions are provided in papers by Bornmann et al. (1994a, 1994b), and Wheatland (2004).

Table content

• Table 2 : Likelihood (%) on C-, M- and X-flares based on today's McI for a sunspotgroup. This table also gives results for the ascending, maximum and descending phase of solar cycle 23.
• Table 3 : Likelihood (%) that a sunspotgroup with certain McI today will evolve into another McI within the next 24 hours. Obviously, this table can also be used to predict further evolution of the sunspotgroup for several days in advance.
• Table 4 : Likelihood (%) on a C-flare within the next 24 hours based on yesterday’s and today’s McI.
• Table 5 : Likelihood (%) on a M-flare within the next 24 hours based on yesterday’s and today’s McI.
• Table 6 : Likelihood (%) on a X-flare within the next 24 hours based on yesterday’s and today’s McI.

Examples

• 04 June 2007 - NOAA 0960 - McI=Fkc
• Flare prediction for that day in % (table 2): C/M/X = 91/55/15
• Most likely McI-evolution in % (table 3): Fkc (44%), Fki (12%), Ekc (7%)
NOAA 0960 produced 1 strong M-flare that day, and retained Fkc-classification.
• 05 June 2007 - NOAA 0960 - McI=Fkc
• Flare prediction for that day in % (table 2): C/M/X = 91/55/15
• Flare prediction for that day in %, but based on yesterday's and today's McI (tables 4, 5, 6): C/M/X = 91/61/16
• Most likely McI-evolution in % (table 3): Fkc (44%), Fki (12%), Ekc (7%)
NOAA 0960 produced only 2 C-flares that day, and evolved into a Ekc-classification.
• 06 June 2007 - NOAA 0960 - McI=Ekc
• Flare prediction for that day in % (table 2): C/M/X = 91/34/7
• Most likely McI-evolution in % (table 3): Ekc (37%), Eki (11%), Fkc (10%)
• Flare prediction for that day in %, but based on yesterday's and today's McI (tables 4, 5, 6): C/M/X = 92/46/14
NOAA 0960 produced 2 C-flares that day (of which one strong), and evolved into a Esi-classification. Because the group was in decline, the predicted reactivation to Fkc was very unlikely (much smaller than 10%).
• 07 June 2007 - NOAA 0960 - McI=Esi
• Flare prediction for that day in % (table 2): C/M/X = 48/2/0
• Most likely McI-evolution in % (table 3): Eso (14%), Eai (13%), Eao (12%), Esi (10%)
• Flare prediction for that day in %, but based on yesterday's and today's McI (tables 4, 5, 6): This McI-evolution apparently had not taken place yet during SC23, hence no predictions can be made based on tables 4, 5, and 6.
NOAA 0960 produced 2 C-flares, and stayed Esi.

It may be clear from these examples that predictions can be improved by attributing the correct McI to the sunspotgroup. From SOHO-images available at the Solar Monitor website, the right McI for NOAA 0960 from 04 to 08 June 2007 seem rather Fkc, Ekc, Esc (Esi?), Esi, Esi. It may be clear that timing of the images need to be taken into account. The flare prediction success-rate can further be improved by taking into account imagery from magnetograms (delta!), H-alpha (filaments!), EIT (brightness) and X-ray (sigmoids!), as well as from other parameters like Wolfnumber, sunspot area, growth/decline rate, and 10,7 cm radioflux.

Original: 09 June 2007