AI-Enhanced Model Could Revolutionize Space Weather Forecasting
A new collaboration between researchers at Los Alamos National Laboratory and the University of North Carolina at Chapel Hill has unveiled a machine learning model designed to enhance space weather forecasting.
This collaboration aims to mitigate the threats posed by “killer electrons” that move at nearly light speed within Earth’s Van Allen belts, a region that traps high-energy charged particles. These electrons can disrupt and damage satellite electronics, posing serious risks to space-borne infrastructure.
The team’s forecasting tool, known as Predictive MeV Electron — Medium Earth Orbit (PreMevE-MEO), offers precise and efficient hourly predictions of electron dynamics within Earth’s outer radiation belt.
By integrating data from 12 GPS satellites in medium-Earth orbit and one Los Alamos satellite in geosynchronous orbit, the model leverages artificial intelligence to significantly improve the accuracy of space weather predictions.
“This study proves the feasibility of using the Laboratory’s particle data to predict the dynamics of killer electrons,” said Yue Chen, a physicist at Los Alamos and the lead author of the study. “It also highlights the importance of long-term space observations in the age of AI.”
A New Era in Space Weather Forecasting
Published in the journal Space Weather, the research marks a significant leap forward in protecting satellites and other space assets. The PreMevE-MEO model utilizes a sophisticated machine learning algorithm that combines convolutional neural networks with transformers, enabling high-fidelity predictions based on decades of satellite observations.
This approach is especially crucial for medium Earth orbit, where many navigation and weather satellites operate. The model’s unique advantage lies in its use of Los Alamos’s extensive dataset, which includes X-ray dosimeter particle measurements.
These datasets, first made publicly accessible in 2017, are archived by the National Oceanic and Atmospheric Administration’s (NOAA) National Centers for Environmental Information. Unlike traditional research missions, which are often short-term, this data represents over 100 satellite years of continuous observations, making it a rich resource for modern AI applications.
Supporting National Space Weather Strategy
The development of PreMevE-MEO aligns with the goals outlined in the recent Implementation Plan for the National Space Weather Strategy and Action Plan. This plan calls for agencies to release historical satellite data, as well as ground-based observatory and magnetometer measurements, to advance the development and validation of space weather models.
By utilizing big data and AI techniques, the Los Alamos team is paving the way for improved space weather forecasting tools that could become operational warning systems. As space becomes increasingly critical for communication, navigation, and national security, enhanced forecasting capabilities are more essential than ever.
