IBM and NASA have joined hands to develop an advanced artificial intelligence-based model called “Soria”. That can predict solar flares and flares produced during solar storms with 16 percent more accuracy. And that too at half the speed of the current system. The AI model is trained on nine years of data from NASA’s Solar Dynamics Observatory (SDO) which takes detailed images of the Sun every 12 seconds. This provides information related to the temperature of the Sun’s layers and the magnetic field. So it is essential for understanding the strength and speed of solar storms.
Experts have transformed this data into a “digital twin” a virtual and evolving image of the Sun that automatically updates as new data arrives. And can be easily observed and analyzed. The model learns from the raw data without complex labeling. Thus making it quickly ready for a variety of tasks, such as predicting solar wind speeds or radiation activity.
In initial experiments, the “Soria” model provided
two hours of advance warning of solar flares, compared to one hour, and increased prediction accuracy by 16 percent, a significant improvement.
“This model makes it easier to understand the complex motions of the Sun with speed and clarity, and it opens up doors for understanding the impacts on future planetary systems,” said NASA Director of Data Science Kevin Murphy.
Solar storms are not only a threat to astronauts, but also severely affect ground communication systems, satellites, power grids, internet networks, and GPS. With this model, the whole defense system can take timely protective measures by receiving a two-hour advance alert. The project’s lead scientist said that “we want to give Earth as much warning time as possible”. So that humans can respond on time when needed. This model could open up new avenues in heliophysics, as well as applications in planetary science or Earth observation. This development also opens new door for the protection of our technology on Earth.
