Researchers from the Los Alamos National Laboratory and the University of North Carolina have developed an artificial intelligence-enhanced space weather forecasting capability.
The LANL said Friday the Predictive MeV Electron – Medium Earth Orbit, or PreMevE-MEO, leverages machine learning to enhance its predictive capabilities. The project, titled “PreMevE-MEO: Predicting Ultra-Relativistic Electrons Using Observations From GPS Satellites,” is intended to enable more accurate hourly forecasts and protect equipment in space.
Machine-Learning Algorithm vs ‘Killer Electrons’
The LANL-UNC collaboration was able to develop a machine-learning algorithm that combined convolutional neural networks with a transformer, allowing the predictive model to measure electrons inside the Earth’s outer radiation belt. These “killer electrons” inside the Van Allen belts can cause malfunctions in space equipment’s electronics.
The PreMevE-MEO was able to observe electrons by utilizing 12 medium-Earth-orbit GPS satellites, and one Los Alamos geosynchronous-Earth-orbit satellite. This means there is potential for predicting space weather based on observations from space infrastructure in medium Earth orbit.
The project, funded by the U.S. Department of Energy and the Laboratory Directed Research and Development program, aligns with the National Space Weather Strategy and Action Plan, intended to enhance preparedness against space weather events.
“This study proves the feasibility of using the Laboratory’s particle data to predict the dynamics of killer electrons,” said Yue Chen, a Los Alamos physicist and lead author of the research. “Meanwhile, it showcases the significance of long-term space observations in the AI age.”