Astrophysicists and data scientists from NASA's COFFIES DRIVE Science Center developed a sliding-window transformer machine-learning model capable of predicting the emergence of storm-causing active sunspot regions on the Sun up to 12 hours before they appear.

Key Details

• Multidisciplinary researchers from NJIT, Princeton University, and NASA Ames Research Center collaborated under the COFFIES DRIVE Science Center
• The team utilized sliding-window transformer architecture to detect subtle precursor reductions in the Sun's acoustic activity and magnetic field
• The model analyzes data captured by NASA's Solar Dynamics Observatory using NASA Ames supercomputing resources
• The findings were published in the Journal of Geophysical Research: Machine Learning and Computation
• Early solar storm prediction capabilities aim to protect astronauts and equipment on NASA Artemis lunar missions and future Mars expeditions

In This Story