AI model predicts solar storms 9.24 hours ahead using acoustic and magnetic data

A research team from the New Jersey Institute of Technology (NJIT) has developed a machine learning model named EarlyDetect, capable of identifying early signs of solar storms up to 9.24 hours in advance. The model analyzes acoustic waves and magnetic field changes on the Sun to detect precursors of active region formation. EarlyDetect uses a Transformer architecture to process data from the Helioseismic and Magnetic Imager (HMI) on NASA's Solar Dynamics Observatory (SDO). HMI records solar vibrations every 45 seconds, which are used to create acoustic power maps. These maps are combined with magnetic field measurements to predict intensity changes. The model’s best version can detect signals indicating an active region is forming 9.24 hours before it becomes visible. The researchers used the solar active region AR11158 as a case study, observing magnetic fields, intensity, and acoustic power changes. They noted that acoustic power decreases first, followed by changes in intensity and magnetic fields. This suggests that sound waves may leave traces before the Sun’s magnetic field rises. If validated, the model could help satellite communication companies and power grid operators prepare for solar storms. However, the team emphasized that EarlyDetect is not yet operational. The study highlights the potential of AI in space weather forecasting, offering a new tool to monitor solar activity. The researchers divided the target area into smaller blocks to track changes continuously. Their findings indicate that acoustic power changes may precede visible active regions, providing an early warning signal. This could allow for better risk assessment and operational adjustments. The team’s work represents a significant step forward in understanding solar dynamics and improving space weather prediction. The model’s success in detecting early signs of solar storms underscores the growing role of AI in scientific research and practical applications.