Solar flares may be best-known as sci-fi MacGuffins, but those flares – and other space weather – can have serious impacts on not only spacecraft and satellites, but also on Earth-based systems such as radio communications and GPS navigation. Accurately monitoring space weather, however, can be difficult, expensive and unreliable. Now, NASA’s Frontier Development Lab (FDL) has developed a new method, based on deep learning, that allows researchers to more accurately capture the Sun’s extreme ultraviolet irradiance.
This extreme ultraviolet irradiance (or EUV) — which is responsible for solar flares — has, in the past, typically been monitored by recording the distribution of magnetic fields or plasma on the Sun and feeding that data into a physics-based model to predict EUV emission.
The FDL, however, used a different approach: researchers created a proxy measurement of EUV irradiance by applying deep learning to the images of the Sun. The images of the Sun are provided by the Solar Dynamics Observatory, a spacecraft that was launched by NASA in 2010 to observe the Sun from geosynchronous orbit around the Earth.
“Our research shows how a deep neural network can be trained to mimic an instrument on the Solar Dynamics Observatory (SDO),” said team member and co-author Alexander Szenicer. “By inferring what ultraviolet radiation levels that sensor would have detected based on what the other instruments on SDO are observing at any given time, we demonstrate it is possible to increase scientific productivity of NASA missions and to increase our capability to monitor solar sources of space weather.”
The results of the deep learning approach surpass current physics-based models – and, as part of the process of evaluating their results, the team developed new benchmarks for comparing predictions, which they hope will come in handy for future studies.
The research is the result of an eight-week summer research accelerator at FDL, which is a partnership between NASA Ames Research Center, the SETI Institute and a number of other private and public partners, including Google Cloud, Intel AI, IBM, Nvidia, HPE and more. This last summer, the challenge was to develop an AI model to use SDO images to predict spectral irradiance.
The breakthrough has been published in Vol. 5, Issue 10 of Science Advances as “A deep learning virtual instrument for monitoring extreme UV solar spectral irradiance.” It was written by Alexandre Szenicer, David F. Fouhey, Andres Munoz-Jaramillo, Paul J. Wright, Rajat Thomas, Richard Galvez, Meng Jin and Mark C.M. Cheung. It can be accessed here. To read the original article from SETI discussing the research, follow this link.