UC scientists use machine learning and AI to predict solar flares

A research team from the Faculty of Sciences and Technology of the University of Coimbra is resorting to machine learning and artificial intelligence (AI) to forecast solar flares – sudden (and so far, unpredictable) intense bursts of energy and radiation from the sun's surface.

This international research is being carried out in the framework of the project SWATNet - Space Weather Awareness Training Network, with the development of a database with several open-access database of various solar phenomena, and the results are quite promising.

According to Teresa Barata, a researcher at FCTUC's Department of Earth Sciences (DCT) and the UC Institute of Astrophysics and Space Sciences (IA), SWATNet’s main goal is to make significant progress in the understanding of the key factors that shape space weather on Earth, using the Marie Skłodowska-Curie Innovative Training Network (MSCA ITN) for the training of early-career researchers in the field of heliophysics.

There are 12 PhD students involved in the project covering different areas of space weather (interaction between the Sun and the Earth), and one of the FCTUC students is getting results in identifying pre-flares; According to the project coordinator at the UC, "identifying pre-flares is crucial as it allows us to develop prediction models and issue timely warnings".

"Solar flares are exerting an increasing impact on Earth. It's not that the sun has changed, but our reliance on space technology has grown significantly with advancements in technology. Consequently, understanding how the sun operates becomes crucial for our safety," warns the expert, further explaining, " Flares are solar eruptions that can reach Earth in as little as eight minutes, carrying energetic particles that can affect our planet."

Neural networks have been used since the late 1990s to predict solar wind speeds or the onset of geomagnetic storms, albeit with limited parameters. "The abundance of available solar observations and recent significant advances in AI and machine learning provide unique opportunities for this study, opening new horizons for the application of these approaches to space weather," says researcher Teresa Barata.

SWATNet offers the 12 participating students a comprehensive experience, including a one-month training course at a solar observatory in Hungary, an exchange programme between institutions and internships in partner companies. In Portugal, the partners include the Pedro Nunes Institute (IPN) and ESA Space Solutions Portugal. The project is also an opportunity to gain insight into the working and business environment.

The project consortium comprises several institutions from countries such as Belgium, Finland, Greece, Hungary, Italy, Poland, Portugal and the UK, as well as several recognised companies in the field.

Further information is available on the project website.