Solar Storm Prediction: New AI & Research for Space Weather Alerts

The ability to predict space weather events – from disruptive solar flares to damaging solar wind – is becoming increasingly crucial as our reliance on satellite infrastructure and terrestrial power grids grows. Recent advancements in artificial intelligence are offering a significant leap forward in forecasting these events, providing potentially days of advance warning where previously only hours were available.

Researchers at New York University Abu Dhabi (NYUAD) have developed an AI model that forecasts solar wind speeds with 45% greater accuracy than current methods, offering a prediction window of up to four days. This breakthrough, detailed in a report from September 17, 2025, centers around analyzing high-resolution ultraviolet (UV) images of the Sun captured by NASA’s Solar Dynamics Observatory. Unlike traditional methods, and even many contemporary AI approaches focused on natural language processing, this system focuses directly on visual patterns within solar imagery.

The impetus for improved forecasting is starkly illustrated by events like the 2022 loss of 40 SpaceX Starlink satellites due to a strong solar wind event. Such incidents highlight the vulnerability of low Earth orbit (LEO) constellations and the broader impact of space weather on critical infrastructure. Solar wind, a continuous stream of charged particles from the Sun, can disrupt Earth’s atmosphere, drag satellites out of orbit, damage their electronics, and even interfere with power grids.

The NYUAD team, led by Postdoctoral Associate Dattaraj Dhuri and Co-Principal Investigator Shravan Hanasoge, trained their AI by correlating UV images with historical records of solar wind data. This approach allows the model to identify subtle visual cues indicative of impending changes in solar wind speed. The system doesn’t rely on interpreting text-based data; instead, it learns directly from the visual characteristics of the Sun.

However, the NYUAD model isn’t the only recent development in this field. NASA and IBM have jointly created Surya, a heliophysics AI foundation model. According to NASA, Surya can generate visual predictions of solar flares two hours into the future, surpassing existing benchmarks by 16%. This model was trained on nine years of observations from the Solar Dynamics Observatory and is designed to help scientists understand solar eruptions and predict space weather’s impact on satellites, power grids, and communication systems.

Surya’s ability to forecast flares is particularly significant. Solar flares are sudden releases of energy from the Sun, often associated with coronal mass ejections (CMEs). CMEs are large expulsions of plasma and magnetic field from the Sun’s corona, and when directed towards Earth, they can cause geomagnetic storms. These storms can disrupt radio communications, damage satellites, and even cause power outages.

Further research, published in Phys.org on February 20, 2026, indicates that new tools are beginning to focus on observing solar active regions – areas on the Sun with intense magnetic activity – to further advance warning times for disruptive solar flares. This suggests a growing trend towards more localized and detailed analysis of the Sun’s surface.

A separate study from New Mexico State University (NMSU) is also contributing to improved flare prediction. Researchers there have developed a new method to predict the occurrence of disruptive solar flares, though details regarding the specific methodology are limited in available reports.

The advancements represented by these AI models – NYUAD’s solar wind forecaster, NASA/IBM’s Surya, and the NMSU flare prediction method – mark a shift towards proactive space weather management. Historically, space weather forecasting has been largely reactive, relying on observations of events already in progress. These new AI-driven systems promise to move the field towards a more predictive capability, allowing operators of critical infrastructure to take preventative measures.

While the current prediction horizons vary – from two hours (Surya for flares) to four days (NYUAD for solar wind) – the trend is clear. The increasing accuracy and lead time offered by these AI models are crucial steps in mitigating the risks posed by space weather. As our dependence on space-based assets and resilient power grids continues to grow, the ability to anticipate and prepare for these events will become even more vital.

The focus on UV imagery, as employed by the NYUAD team, is particularly noteworthy. Ultraviolet light reveals details about the Sun’s corona that are not visible in other wavelengths, providing valuable insights into the processes that drive solar activity. The success of this approach suggests that further investment in high-resolution UV imaging could yield even more accurate and timely space weather forecasts.