Prediction for 4 November 2003 

On 4 November 2003 at 19:29UT the active region designated 10486
produced the largest
soft Xray flare of the modern era. For details, see e.g.
here.
This flare was classified as an X28 event by the NOAA, i.e. the flare
had a peak soft Xray 0.0028 Watts per square metre in the 18 Angstrom
band of the GOES satellites, although this was an
estimate because the flare saturated the detectors on the satellites.
Based on the effect of the flare on the Earth's ionosphere, Thomson, Rodger
and Dowden (2004, Geophysical Research Letters 31, L06803) estimate that
the event was actually X45!
It is of interest to try the prediction method for that day (after the fact). The prediction, based on one year of data prior to 4 November 2003, is as follows.

Prediction made at: 20031104T00:00:00.000 for events within 1.00 days based on 365.25 days of data THRESH = 4.0e06 Number of flares in year of data = 480 Maximum likelihood GAMMA = 2.069+/0.049 Total number of events= 480 Segmentation procedure complete  calculating rates... Check on moments for prior = 1.2e08 Number of blocks = 13 Number of events in last block = 104 Duration of last block (days) = 15.30 Prob. of at least one flare of size M to X = 0.725+/0.027 Prob. of at least one flare of X size or greater = 0.192+/0.017 
So the method identifies a 73 percent chance of at least one event of size M to X, and a 19 percent chance of at least one X class event. These are relatively high predictions for the method, so in some sense it has "succeeded". However, this example also highlights the limitations of probabilistic forecasting: the forecast for an X class flare is only 19%, yet the biggest flare of the modern era is imminent. More details on the prediction are provided by the following three figures. (For explanations of these figures, see this page, with the corresponding figures for today's prediction.)

Figure 1: Top panel: Year of flaring used in prediction. Bottom: Bayesian blocks decomposition.

Figure 2: Top panel: Differential distribution of year of events in peak flux. Bottom: Cumulative distribution in peak flux.

Figure 3: Top panel: Posterior distribution for MX events. Bottom: Posterior distribution for X events. 

Main flare prediction page
More detail on how today's predictions were made Summary of all predictions made to date Automatic check of the reliability of the predictions A test of the method on historical data Links to other pages related to flare prediction Mike Wheatland's home page 


Page maintained by m.wheatland@physics.usyd.edu.au  Page last updated Wednesday, 01Dec2004 15:39:19 AEDT 