Machine Learning Model Runs a Million Times Faster to Forecast Space Weather
-
Researchers created a machine learning model, SMRAI2, to emulate complex physics-based simulations of the auroral current system caused by solar storms.
-
SMRAI2 runs a million times faster than physics simulations and can be run on a laptop rather than requiring a supercomputer.
-
The model incorporates seasonal effects and is the first ML emulator of physics-based ionospheric simulations.
-
Accurate space weather forecasts help communities prepare for storms' effects on satellites, GPS, communications, electrical grids, and radiation exposure.
-
The researchers plan to incorporate the emulator into ensemble and data assimilation forecasts to further improve prediction accuracy.