The intricate enigmas of the Sun may soon be deciphered with assistance from artificial intelligence. On August 20, IBM and NASA introduced Surya, a foundational model tailored for the sun. This AI instrument, trained on vast datasets of solar phenomena, seeks to improve our comprehension of solar weather and forecast solar flares—outbursts of electromagnetic radiation from the sun that threaten astronauts and disrupt Earth’s communication systems.
Surya was crafted using nine years of information from NASA’s Solar Dynamics Observatory (SDO), which has circled the sun since 2010, capturing high-resolution visuals every 12 seconds. The SDO offers observations across various electromagnetic wavelengths to assess the sun’s temperature and gathers accurate measurements of its magnetic field—crucial data for understanding energy movement and forecasting solar storms.
Historically, interpreting this extensive and intricate data has posed difficulties for heliophysicists. To address this, IBM utilized SDO data to create a digital duplicate of the sun—a dynamic virtual model refreshed with new information for easier examination.
The initial step involved standardizing the assorted data formats for uniform processing. An advanced long-range vision transformer, an AI framework, was later utilized to examine high-resolution images and discern connections between components, irrespective of distance.
The model’s efficacy was augmented using spectral gating, which diminishes memory consumption by up to 5 percent by eliminating noise, thereby enhancing data quality.
More Precise Predictions in Shorter Timeframes
Surya’s creators assert it offers a notable edge: it learns directly from unprocessed data, unlike other algorithms that depend on extensive data labeling. This enables Surya to swiftly adapt to various tasks and produce consistent results efficiently.
During evaluations, Surya demonstrated its proficiency in integrating data from other devices, such as the Parker Solar Probe and the Solar and Heliospheric Observatory (SOHO), which also monitor the sun. It was effective in forecasting flare occurrences and solar wind velocities.
IBM indicates that conventional models can foresee a flare one hour ahead based on signals from certain sun regions. Conversely, Surya provided a two-hour advance using visual data and increased solar flare classification accuracy by 16 percent in initial evaluations.
NASA emphasizes that although Surya was crafted for heliophysics, its versatile structure can extend to domains like planetary science and Earth observation. “By establishing a foundational model trained on NASA’s heliophysics data, we’re facilitating the analysis of the sun’s behavioral complexities with unparalleled speed and accuracy,” noted Kevin Murphy, NASA’s director of data science. “This model enables a broader comprehension of how solar activity influences essential systems and technologies that we all depend on here on Earth.”
Unusual solar activity presents serious hazards, potentially impacting global telecommunications, crippling electrical grids, and disrupting GPS navigation, satellite operations, internet connections, and radio communications.
Andrés Muñoz-Jaramillo, a solar physicist at the Southwest Research Institute and the project’s lead scientist, emphasized that Surya aims to maximize lead time for such occurrences. “Our goal is to provide Earth with the longest lead time possible. We hope that the model has comprehended all the crucial processes behind our star’s evolution over time, allowing us to derive actionable insights.”
This story originally appeared on WIRED en Español and has been translated from Spanish.
