Reviewed by Lexie CornerNov 8 2024
Researchers at Los Alamos National Laboratory and the University of North Carolina at Chapel Hill are collaborating to improve a predictive model for measuring electrons within the Earth's outer radiation belt using machine learning, a form of artificial intelligence. This research is detailed in a study published in Space Weather.
Electrons travel at nearly light speed inside Earth's Van Allen belts and pose a major threat to satellites and space equipment. Image Credit: Los Alamos National Laboratory
By causing electronic malfunctions, “killer electrons,” which move at almost the speed of light inside Earth’s Van Allen belts—the region that envelops the planet and traps energetic charged particles—pose a serious risk to equipment in space.
This study proves the feasibility of using the Laboratory’s particle data to predict the dynamics of killer electrons. Meanwhile, it showcases the significance of long-term space observations in the AI age.
Yue Chen, Study Lead Author and Physicist, Los Alamos National Laboratory
The study marks a significant advancement in satellite protection and space weather forecasting capabilities.
The team has developed a new forecasting tool called Predictive MeV Electron—Medium Earth Orbit (PreMevE-MEO), which can produce more precise and effective hourly forecasts.
PreMevE-MEO utilizes observations of electrons from one Los Alamos geosynchronous Earth orbit satellite and twelve medium Earth orbit GPS satellites as inputs for the model. To enhance its performance, the researchers created a novel machine-learning algorithm that integrates transformers and convolutional neural networks.
As a result, the team demonstrated that high-fidelity predictions can be made based on data from established space infrastructure operating in medium Earth orbit, the altitude at which many navigation and weather satellites function. This model has the potential to evolve into an efficient operational warning system for space weather.
The National Oceanic and Atmospheric Administration's National Centers for Environmental Information has archived X-ray dosimeter particle data, which was first made publicly available in 2017. This data has been integrated with unique GPS data from Los Alamos in the new model.
A notable aspect of this dataset is that it represents a long-term constellation with over 100 satellite years of data, distinguishing it from conventional NASA research missions. This resource is one of the few from the space environment that qualifies as big data and can be utilized with modern AI techniques.
This study aligns with the recent National Space Weather Strategy and Action Plan Implementation Plan, which instructs agencies to identify and make accessible historical data from satellites, ground-based observatories, and networks funded by the U.S. government. Such data includes measurements across the electric power grid and magnetometer data streams that contribute to the development, validation, and testing of models used to describe and predict space weather events.
The U.S. Department of Energy and the Laboratory Directed Research and Development program funded the research under Grant Number 20230786ER.
Journal Reference:
Feng, Y. et. al. (2024) PreMevE-MEO: Predicting Ultra-Relativistic Electrons Using Observations From GPS Satellites. Space Weather. doi.org/10.1029/2024SW003975