This product shows the prediction of a flare occurrence within the next 24 hours after the given date,
and starting at 3AM. The probability is based on the best available machine learning algorithm tested in
FLARECAST, Random Forest. The flare size assumed is either C, M, or X, following the GOES classification.
The NOAA active region number(s) is or are indicated, and the diagram in the grey area emphasizes the
different probability for each active region. If there are no active regions present, the system cannot
make any prediction and issues a corresponding message.
The FLARECAST consortium developed an automated forecasting system for solar flares. The team integrated
virtually every solar flare-predicting parameter into an open online application programming interface,
flexible enough to facilitate future expansion. We identified the best performers by employing a variety of
statistical and machine learning techinques, including standard methods such as Linear Discriminant
Analysis, Clustering and Regression Analysis, Neural Networks, as well as innovative approches including
Multi-Task Lasso, Simulated Annealing and Random Forest. A robust exploration work package identified
promising new predictors and connected flare prediction to other manifestations of solar eruptive activity
such as coronal mass ejections.
For more information please visit this page:
A simple REST API gives access to the FLARECAST flare forecasts of a whole year. You may use any
common HTTP client such as
Make sure to pass a valid authentication cookie to the method.
Year: four digit year after 2013 or