NASA, IBM Release Heliophysics AI Foundation Model Trained on Solar Data

NASA’s Office of the Chief Science Data Officer is creating a suite of open-source artificial intelligence foundation models using data from each of NASA’s five science divisions, allowing researchers to more easily conduct AI-powered scientific studies. The latest addition to this suite, created in collaboration with IBM and publicly released on Hugging Face on August 20, is the Surya Heliophysics Foundation Model.

Surya was trained on nine years of observations from NASA’s Solar Dynamics Observatory. Initial results show that Surya can predict solar flares two hours into the future with an accuracy 16% above the benchmark. A higher accuracy in sensing solar activity buys decision-makers vital time to prevent solar flares from affecting electronic systems on Earth and harming astronauts in deep space.

The model was co-created by the Interagency Implementation and Advanced Concepts Team (IMPACT) under NASA’s Office of Data Science and Informatics (ODSI) at NASA’s Marshall Space Flight Center in Huntsville, Alabama.

To learn more about Surya and its preliminary results, read the recent web feature. To learn more about the team and technology behind Surya, read the latest post on the Science Data Portal blog.

BurstCube Public Data Archive Opens

BurstCube was a shoebox-sized NASA CubeSat mission that detected gamma-ray bursts — powerful explosions from distant galaxies. The mission collected data from May 2024 through August 2024. Now, all the datasets collected by BurstCube are openly available through NASA’s High Energy Astrophysics Science Archive Research Center (HEASARC).

For more information about BurstCube, its science goals, and its data, visit the BurstCube web page.

First PREFIRE CubeSat Data Released

NASA’S PREFIRE (Polar Radiant Energy in the Far-InfraRed Experiment) mission, consisting of a pair of tiny CubeSats, aims to capture the first comprehensive measurements of far-infrared (FIR) radiation in the Arctic and Antarctic regions. Beginning in August, PREFIRE’s Level 1, 2, and 3 datasets became available in a public archive. These data products include space-based measurements of radiative fluxes, cloud masks, spectrally variant surface emissivity, and column water vapor.

Data from PREFIRE will help reveal the complete spectrum of radiant energy in the polar regions, improving our understanding of Earth's energy budget and the role FIR radiation plays in changes to the land, ice, and ocean near the poles. For more information, read the news article on the Earthdata website.

Physical Science Informatics Adds 4 New Datasets

NASA’s Physical Sciences Informatics (PSI) data repository published five new open datasets in August, revealing breakthroughs in materials science and helping improve the accuracy of microgravity experiments:

PSI-185: This ground study utilized flight data found in PSI to test models of phase coarsening in solid-liquid mixtures to better understand how materials change over time. The research results can help improve engineering designs and make tools and materials that last longer.

PSI-186: Conducted during the Space Shuttle Life and Microgravity Spacelab (LMS) mission, this materials science investigation demonstrated, for the first time, steady-state fibrous coupled growth in immiscible alloy systems.

PSI-181: Using existing flight data found in PSI from the In-Space Soldering Investigation (ISSI), this PSI grant project modeled the structural and mechanical behavior of solder alloys under terrestrial and microgravity conditions. Its analysis of how defects form in solder joints under microgravity and extreme thermal conditions will help improve spacecraft and equipment for long-duration space missions.

PSI-177: This flight investigation explored how particle shape and toggling magnetic fields affect the separation and self-organization of particles in fluids, both on the International Space Station and on Earth. The results revealed the potential for new methods of controlling material assembly in Earth-based manufacturing.

PSI-166: In partnership with the European Space Agency (ESA), this research used magnetohydrodynamic simulation to improve experimental parameters across various experiments to be performed on the International Space Station Electromagnetic Levitator (ISS-EML). This investigation led to new discoveries and provided new insights in dynamic nucleation while challenging some classical nucleation theory.

To explore all openly available PSI datasets, visit the PSI repository.