NASA and IBM Launch Surya A New AI Model for Space Weather Forecasting

NASA, in partnership with IBM, has introduced Surya a new avant-garde open-source artificial intelligence model to transform the forecast of dangerous space weather.

Why Prediction is needed

The Solar eruptions such as flares and coronal mass ejections (CME) form what is known as Space weather, which is a major threat to the modern technology. Clashes can destroy satellites, trip up power grid, interfere with GPS and air travel systems and also put astronauts at risk. Proper forecasting is an important aspect of curbing such risks.

The Surya Transformation in some ways Surya is transforming forecasting How Surya is transforming forecasting Surya was trained on nine years of records of NASA Solar Dynamics Observatory data set and uses powerful machine learning to recognise small parcels of information in the activity of the sun. This will enable it to accurately predict solar eruptions in a way never before seen in the world and will supply lead times of two or more hours, which is also large compared to what is currently available.

Technical Breakthrough

Insurmounting a longstanding difficulty with solar modeling, Surya is able to capture multi-scale activity on the Sun. It allows a new distinctive architecture by mixing spectral spectrum block layers and a long-short transformer backbone. This allows the model to evaluate both the wider and narrowest of the details and give a more in-depth and correct analysis.

Proven Capabilities

Surya has already proven to be of value in testing. It has been able to re-create the historic St Patrick day 2015 geomagnetic storm. The model is successful in four major functions namely predicting a solar flare, forecasting an emergence of active areas, velocity of solar wind and the extreme ultraviolet spectra. It performed better by up to 16 percent than the existing models in predicting flares.

A Co-Creative, Open Source Future

Surya was developed by an alliance of NASA workers, universities and industry partners with assistance of the National Science Foundation and NVIDIA as one of the first attempts at AI-assigned heliophysics. Being open-source, NASA allows other, global researchers to add to its development, making the platform more innovative and resistant to space weather threats on a collective level.