Solar wind, a constant stream of charged particles from the Sun, can intensify into powerful events that disrupt Earth's atmosphere, interfere with power grids, damage satellites, and even push spacecraft out of orbit. A dramatic example occurred in 2022, when a solar storm destroyed 40 newly launched Starlink satellites, underlining the urgent need for reliable forecasting.
The NYUAD research team, led by postdoctoral associate Dattaraj Dhuri and co-principal investigator Shravan Hanasoge at the Center for Space Science, developed a neural network trained on high-resolution ultraviolet imagery from NASA's Solar Dynamics Observatory alongside decades of solar wind data. Unlike popular AI language models that interpret text, their system interprets images of the Sun to recognize patterns that precede solar wind variations.
The model achieved a 45 percent boost in predictive accuracy over operational forecasting systems and improved results by 20 percent compared with previous AI-based techniques. "This is a major step forward in protecting the satellites, navigation systems, and power infrastructure that modern life depends on," Dhuri said. "By combining advanced AI with solar observations, we can give early warnings that help safeguard critical technology on Earth and in space."
This advance shows how machine learning can address one of astrophysics' most difficult problems: anticipating the solar wind. Stronger predictive power enables governments, researchers, and industries to prepare for disruptive space weather, protecting essential technology and services worldwide.
Research Report:A Multimodal Encoder - Decoder Neural Network for Forecasting Solar Wind Speed at L1
Related Links
New York University
Solar Science News at SpaceDaily
| Subscribe Free To Our Daily Newsletters |
| Subscribe Free To Our Daily Newsletters |
