To the casual observer, the Sun seems to be the one constant and never changing. The reality is that the Sun is a seething mass of plasma, electrically charged gas which is constantly being effected by the Sun’s magnetic field. The unpredictability of the activity on the Sun is one of the challenges that faces modern solar physicists. The impact of coronal mass ejections are one particular aspect that comes with levels of uncertainty of their impact. But machine learning algorithms could perhaps have given us more warning! A new paper suggests algorithms trained on decades of solar activity saw all the signs of increased activity from the region called AR13664 and perhaps can help with future outbursts.
Coronal Mass Ejections or CMEs, are massive bursts of plasma ejected from the Sun’s corona into space due to disruptions in the Sun’s magnetic field. These explosive events are often linked to flares and occur when magnetic field lines suddenly realign, releasing vast amounts of energy. CMEs can travel at speeds ranging from a few hundred to several thousand kilometres per second, sometimes reaching Earth within days, if their trajectory is in our direction. When they arrive, they can interact with our magnetosphere and trigger geomagnetic storms, potentially disrupting satellite communications, GPS systems, and power grids. Additionally, they can lead to auroral activity, creating breathtaking displays of the northern and southern lights.
Accurately forecasting these types of events and how they impact our magnetosphere has been one of the challenges facing astronomers. In a study authored by a team of astronomers led by Sabrina Guastavino from the University of Genoa, they applied artificial intelligence to the challenge. They used the new technology to predict the events that were associated with the May 2024 storm, the corresponding flares from the region designated 13644 and CMEs. The storm unleashed intense solar events including a flare classed as an X8.7!
Using AI the team were able to point machine learning technology to the vast amounts of previously collected data to uncover complex patterns that were not easy to spot using conventional techniques. The 2024 event was a great, and unusual opportunity to test the AI capability to predict solar activity. The chief objective was to predict the occurrence of solar flares, at how they changed over time, CME production and ultimately, to predict geomagnetic storms here on Earth.
They ran the process against the May 2024 event with impressive results. According to their paper, the prediction revealed ‘unprecedented accuracy in the forecast with significant reduction in uncertainties with respect to traditional methods.’ The results of the CME travel times to Earth and the onset of geomagnetic storms was also impressively accurate.
The impact of the study is profound. Power grid outages, communication and satellite issues can be a major disadvantage when CMEs hit Earth so the application of the machine learning AI toolset to predicting solar activity looks like an exciting advance. For those of us keen sky watchers, we may also get a far better forecast of auroral activity too.
