China-led team develops AI system to track radar-disrupting space hurricanes
A China-led team has developed an AI system to detect space hurricanes, a phenomenon in the Earth’s upper atmosphere that can disrupt satellite signals, radar and radio communications. While the hurricane-like atmospheric phenomenon can have major space weather effects, detection has so far relied o
By Victoria Bela
A China-led team has developed an AI system to detect space hurricanes, a phenomenon in the Earth’s upper atmosphere that can disrupt satellite signals, radar and radio communications.
While the hurricane-like atmospheric phenomenon can have major space weather effects, detection has so far relied on a tedious process of studying satellite images manually.
The team said it had developed a new deep-learning system that could automatically detect and pinpoint space hurricanes through ultraviolet images, which they said could be used to analyse data from a newly launched China-Europe satellite.
“A space hurricane is a recently discovered space weather event that appears as a massive, spinning aurora near Earth’s magnetic poles,” the team said in a paper published in the peer-reviewed journal Space Weather on May 23.
The phenomenon is named after the tropical cyclones that occur in the north Atlantic and northeastern Pacific, which are the same weather phenomena as typhoons in the northwestern Pacific.
Instead of clouds of rain and wind, space hurricanes are swirling masses of plasma that pour high-energy electrons into the polar ionosphere – a highly charged region within the Earth’s upper atmosphere – generating a massive, cyclone-shaped aurora.
The first documented event was a 1,000km-wide (620-mile) spiral of plasma that swirled above the North Pole for around eight hours in 2014. The event was only proved and published in a Nature Communications paper by a Shandong University-led team in 2021.
Since then, scientists have found that these storms in the Earth’s ionosphere and magnetosphere can disrupt radio communication, cause navigation and positioning errors, and impair over-the-horizon radar detection.
The team, led by researchers from Shandong University and the Chinese Academy of Sciences’ National Space Science Centre, said the present standard of manually inspecting space hurricanes was subjective, inefficient and lacked the tools for automatic recognition.
“To overcome this, we developed an artificial intelligence system that can automatically spot and pinpoint space hurricanes in ultraviolet images from satellites,” the researchers said in their paper.
The team said the best model from its deep learning architecture could achieve nearly 98 per cent accuracy in detecting space hurricanes, “significantly outperforming previous approaches”.
According to the authors, the identification method could be used to help study ultraviolet images from the Solar Wind Magnetosphere Ionosphere Link Explorer (SMILE) satellite, a joint project between China and Europe, which launched in May.
During a space hurricane, they said, the energy deposition that could occur at the polar ionosphere was comparable to intense magnetic storms.
“The establishment of an intelligent space hurricane identification and monitoring framework therefore offers considerable scientific value, enhancing capabilities for monitoring space environment hazards and supporting the safety of polar communications and aviation,” the team said.
To build the system, the team used auroral observations and particle detection data from a sensor aboard Defence Meteorological Satellite Programme polar-orbiting satellites, which were launched by the US Air Force to monitor the Earth’s space environment.
Using 300,000 auroral images collected from 2005 to 2021 in both hemispheres, the team selected 570 distinct space hurricane events, which they combined with samples that were not space hurricanes, including some that resembled them.
This large data set was used to train several advanced computer vision models, with one able to locate the space hurricanes precisely.
“We also created a complete software system with a visual interface to simplify this detection process for researchers,” the team said.
“This work provides a fast, reliable tool that helps advance the study of space weather in polar regions and supports the development of early warning systems for related technological disturbances.”
The researchers said their methodology was “well positioned” to support space missions such as SMILE – which was developed by the Chinese Academy of Sciences and the European Space Agency – because it had an ultraviolet imager to capture continuous, high-resolution auroral images.
“The efficient and automated detection of space hurricanes – as demonstrated in this study – will enable systematic processing of the large volumes of UV imagery expected from SMILE.”
Their next step is to advance towards forecasting space hurricanes, a process that will integrate real-time data sources to establish a space-ground monitoring network to support so-called nowcasting of the immediate future and short-term forecasting.