The aurora borealis, also known as the northern lights, is a captivating natural phenomenon observed across the polar regions. This breathtaking display of colors in the sky occurs when charged particles from the sun interact with Earth’s magnetic field, producing shimmering curtains of light. While this ethereal sight has mesmerized observers for centuries, scientists have also come to recognize that the aurora can have a significant impact beyond its beauty. The same solar activity that causes auroral displays can also lead to geomagnetic storms—powerful disturbances in Earth’s magnetosphere. These storms can wreak havoc on vital infrastructure such as communication satellites, GPS systems, and power grids.
Understanding these disturbances and their impact on Earth’s systems is crucial. Recent advancements in research are helping scientists gain better insights into how geomagnetic storms unfold, thanks in part to the application of artificial intelligence (AI) and machine learning. One significant step forward in this area of research comes from a team at the University of New Hampshire (UNH), which has developed an innovative AI-based methodology to help analyze and understand auroral data.
A groundbreaking study, recently published in the Journal of Geophysical Research: Machine Learning and Computation, introduced a sophisticated algorithm designed to categorize the largest-ever database of aurora images. Researchers have long sought ways to harness the massive volumes of data captured by space instruments that monitor the aurora, and now, AI and machine learning techniques are making this process more efficient and meaningful.
This research leverages data from NASA’s Time History of Events and Macroscale Interactions during Substorms (THEMIS) mission, which involved a pair of spacecraft tasked with studying the space environment around Earth. THEMIS collects high-resolution images of the night sky every three seconds from 23 observation stations spread across North America, providing a continuous view of auroral phenomena. Covering the years from 2008 to 2022, the data sets are massive, containing images that represent one of the most extensive repositories of aurora observations in history.
Jeremiah Johnson, the lead author of the study and an associate professor at UNH’s Department of Applied Engineering and Sciences, emphasizes the immense value of this database. “The massive dataset is a valuable resource that can help researchers understand how the solar wind interacts with the Earth’s magnetosphere,” Johnson explained. The magnetosphere is Earth’s protective shield, deflecting charged particles from the sun and preventing them from directly bombarding our atmosphere. However, during periods of intense solar activity—when large eruptions or coronal mass ejections occur on the sun—these particles can breach this magnetic shield, leading to geomagnetic storms. These storms can then disrupt satellites, global positioning systems (GPS), communication networks, and power grids, sometimes causing widespread technological malfunctions.
Despite the extraordinary potential of the THEMIS dataset, analyzing and interpreting such a massive collection of images proved a challenge. In the past, researchers struggled to efficiently process and sort the data in a meaningful way, as the sheer size of the database made it cumbersome to handle manually. The advent of artificial intelligence and machine learning, however, offers new methods for organizing and analyzing this data at scale. For example, AI tools have made it possible to classify over 706 million images from the THEMIS mission—a task that would have been insurmountable with traditional methods.
By developing a new algorithm, the UNH team has made it easier to categorize the auroral images into distinct categories. They divided the data into six primary categories based on observable auroral phenomena: arc, diffuse, discrete, cloudy, moon, and clear/no aurora. This system of annotation enables researchers to quickly sort and retrieve specific types of aurora-related data, significantly improving the utility of the database. Whether studying the light patterns formed by auroral arcs or assessing images showing periods of cloudy skies that obscure auroras, the categorized data provides critical insights.
But the usefulness of the categorized database extends beyond simple organization. “The labeled database could reveal further insight into auroral dynamics,” said Johnson. A deeper understanding of these dynamics will help researchers predict the occurrence and impact of geomagnetic storms on Earth. By detecting patterns in auroral activity that often precede geomagnetic storms, scientists can refine their forecasting models, improving our ability to predict when and where such storms are likely to occur. This is especially important for industries and organizations that rely on space-based technologies.
In addition to improving geomagnetic storm prediction, the categorized THEMIS database is also invaluable for the broader scientific community. The wealth of historical auroral data will enhance future studies of Earth’s magnetosphere, space weather, and auroral phenomena. Studying how the solar wind affects the magnetosphere and its role in geomagnetic storms could also deepen our understanding of space weather’s broader effects on human technology, climate, and infrastructure.
The research team’s work also included contributions from Amy Keesee, an associate professor of physics and astronomy at UNH, who brought her expertise in space weather, and several collaborators from the University of Alaska–Fairbanks, including Doğacan Su Öztürk, Donald Hampton, and Matthew Blandin. The team also collaborated with Hyunju Connor from NASA’s Goddard Space Flight Center. These co-authors are renowned experts in their respective fields, with each bringing unique skills to help advance the study of auroral dynamics and space weather.
While the new system of sorting and categorizing aurora images is one of the most promising developments for auroral research, there’s still more to learn. “At a very basic level,” Johnson noted, “our goal was to organize the THEMIS all-sky image database so that the vast amount of historical data it contains can be used more effectively by researchers.” However, the labeled dataset opens up a wealth of possibilities for further exploration.
For example, studying the connection between auroral intensity and geomagnetic storm occurrence could lead to earlier warning systems for power companies and satellite operators, protecting these critical industries from damage caused by solar storms. Likewise, investigating auroral behavior in real-time could help researchers gain better insight into long-term space weather patterns that influence Earth’s atmosphere and climate. Understanding such patterns may have implications for future space exploration, especially as missions to the moon, Mars, and beyond become more common. Space weather is a crucial factor in planning these ambitious missions, as magnetic field disturbances could interfere with spacecraft navigation systems or radiation-sensitive instruments.
Reference: Jeremiah W. Johnson et al, Automatic Detection and Classification of Aurora in THEMIS All‐Sky Images, Journal of Geophysical Research: Machine Learning and Computation (2024). DOI: 10.1029/2024JH000292