A new artificial intelligence model called EarlyDetect can spot faint precursor signatures of solar active regions an average of 9.24 hours before they emerge on the Sun’s visible surface, according to research published in the Journal of Geophysical Research: Machine Learning and Computation. Developed by researchers at the New Jersey Institute of Technology, the system aims to solve a long-standing challenge in space weather forecasting by detecting subtle acoustic and magnetic shifts beneath the solar photosphere before sunspots form.
How EarlyDetect Tracks Sub-Surface Solar Activity
Forecasting space weather has historically been hindered by the fact that active regions develop beneath the Sun’s visible surface, where instruments cannot directly observe the underlying magnetic structure. According to Dr. Alexander Kosovichev, a distinguished professor of physics at NJIT and co-author of the study, researchers must instead look for minute changes in the magnetic field and acoustic waves traveling through the star. Kosovichev compared the detection process to finding a slight change in rhythm within a noisy orchestra.
To overcome this observational barrier, the research team built EarlyDetect using a Transformer architecture—the same underlying framework that powers large language models like ChatGPT and Gemini, according to project details. The system analyzes hourly acoustic power maps and magnetic measurements supplied by the Helioseismic and Magnetic Imager onboard NASA’s Solar Dynamics Observatory, commonly referred to as SDO/HMI. By processing these rapid 45-second observational intervals, the AI model successfully identifies declining acoustic power and subsequent magnetic shifts that precede sunspot visibility.
Comparing Machine Learning With Historical Solar Storms
Advanced machine learning models like EarlyDetect represent a shift in how scientists approach solar activity prediction. According to Dr. Mengjia Xu, an assistant professor of data science at NJIT and principal investigator on the project, machine learning has not been widely applied to solar activity forecasting until now, meaning this work opens entirely new possibilities for predictive modeling.
Improved lead times carry significant practical stakes for modern infrastructure. In contrast to modern predictive tools, historical solar storms struck without warning. The most famous space weather event in modern history, the Carrington Event of September 1–2, 1859, generated solar flares equivalent to 10 billion atomic bombs, according to historical accounts. Discovered by British astronomer Richard Carrington, the intense flares caused global telegraph networks to fail, shocked teleoperators, and even ignited telegraph paper within hours, leading contemporaries to speculate that an apocalypse was underway.
Did You Know?
During the Carrington Event of 1859, auroras were visible worldwide, and some telegraph systems continued to operate even after operators had completely disconnected their power supplies because the induced electrical current from the solar storm powered the lines directly.
Preparing Power Grids and Satellites for Solar Storms
Gaining advance notice before solar active regions become apparent could fundamentally change how industries protect critical assets. According to Jonas Tirona, an NJIT undergraduate researcher and corresponding author for the study, an early warning window averaging over nine hours allows satellite communications companies and power grid operators time to prepare and mitigate potential damage from incoming solar storms.
While EarlyDetect is not yet operational, the validation tests demonstrate that automated pattern recognition can isolate faint precursory signatures from heavy background solar noise. As researchers continue refining these neural networks, the capability to anticipate solar eruptions before they break the photosphere moves closer to operational reality.
Frequently Asked Questions
What is EarlyDetect?
EarlyDetect is an artificial intelligence model built on a Transformer architecture designed to analyze acoustic waves and magnetic field data from the Sun to detect precursor signals of solar active regions before they become visible.
How far in advance can EarlyDetect spot solar active regions?
According to findings published in the Journal of Geophysical Research: Machine Learning and Computation, EarlyDetect successfully identified precursor signals an average of 9.24 hours before active regions became visible on the Sun’s surface.
Why are sub-surface solar active regions difficult to study?
Active regions develop beneath the Sun’s visible photosphere, forcing scientists to rely on indirect methods like helioseismology to track acoustic wave patterns and subtle magnetic shifts.
Stay Updated on Space Weather Discoveries
Want to explore more breakthroughs in space science and AI research? Drop a comment below with your thoughts, share this article with fellow science enthusiasts, or subscribe to our newsletter for regular updates.
Related reading