Researchers use AI to help simulate and predict solar events

This could help us improve our understanding of the Sun and its impact on space weather.
Ameya Paleja
Predict solar events accurately is still in its infancy
Accurate prediction of solar events is still in its infancy

Iurii Garmash/iStock 

A collaborative effort between researchers at the University of Graz in Austria and the Skolkovo Institute of Science and Technology (Skoltech) in Russia used artificial intelligence (AI) to study the magnetic field in the upper atmosphere of the Sun, a press release said.

The solar magnetic field is a poorly understood area of research among astronomers. Even after centuries of watching the Sun, we only have limited information about how sunspots are formed or whether they will lead to events like a flare or a coronal mass ejection (CME).

Large amounts of electromagnetic radiation leave the Sun during these events, determining our space weather. As humanity aims to traverse further away from the comforts of our atmosphere, the ability to predict space weather will be a crucial part of our survival in the extremely harsh conditions of outer space.

Our current capabilities to observe the Sun's magnetic field only allows us to see the surface of the Sun. However, the build-up and release of energy happen higher up in the corona, the Sun's atmosphere.

Using AI to simulate Sun's magnetic fields

AI has been leveraged to accelerate the pace of discovery in many fields of research ranging from drug discovery to computer science. Researchers from Graz and Skoltech used a neural network that was trained on data from physics and integrated observational data with a magnetic field model.

Researchers use AI to help simulate and predict solar events
Composite of extreme ultraviolet observation (left) and magnetic field (right). The magnetic field lines are obtained from the simulation and show agreement with the outlined structures in EUV.

Doing so helped the researchers reach a comprehensive understanding of the connections between the phenomena observed and the underlying physics that was occurring on the solar surface.

To determine if the model worked, the researchers then simulated the evolution of a solar active region. The AI model needed just 12 hours of computation time to simulate the turn of events that would happen over the period of the next five days.

Predicting solar events

The team then further used its computational system to study the free magnetic energy within the coronal volume. Events like CMEs are linked to coronal volume, and the observation of ultraviolet radiation confirmed the accuracy of the method. Later on, the researchers observed reductions in free energy in these regions, which were expected results after a solar eruption.

"The use of AI techniques for numerical simulations allows us to better incorporate observational data and holds great potential to further advance our simulation capabilities," said Tatiana Podlachikova, associated professor at Skoltech, in the press release. "Our use of artificial intelligence in this context represents a transformative leap forward."

The rapid pace of computation achieved in these simulations helped the researchers make their predictions in near real time. "The computing speed holds significant promise for improving space weather forecasting and advancing our knowledge of the Sun’s behavior," Podlachikova added.

The research findings were published in the journal Nature Astronomy.

Earlier in March, Interesting Engineering reported that an AI-based prediction model used by NASA could also predict solar storms up to 30 minutes in advance.


While the photospheric magnetic field of our Sun is routinely measured, its extent into the upper atmosphere is typically not accessible by direct observations. Here we present an approach for coronal magnetic-field extrapolation, using a neural network that integrates observational data and the physical force-free magnetic-field model. Our method flexibly finds a trade-off between the observation and force-free magnetic-field assumption, improving the understanding of the connection between the observation and the underlying physics. We utilize meta-learning concepts to simulate the evolution of active region NOAA 11158. Our simulation of 5 days of observations at full cadence (12 minutes) requires less than 12 hours of total computation time, allowing for real-time force-free magnetic-field extrapolations. The application to an analytical magnetic-field solution, a systematic analysis of the time evolution of free magnetic energy and magnetic helicity in the coronal volume, as well as comparison with extreme-ultraviolet observations, demonstrates the validity of our approach. The obtained temporal and spatial depletion of free magnetic energy unambiguously relates to the observed flare activity.

Add Interesting Engineering to your Google News feed.
Add Interesting Engineering to your Google News feed.
message circleSHOW COMMENT (1)chevron
Job Board