AI-driven solar model aims to extend space weather warnings

by Silas Korrin
New York (SDX) Feb 23, 2026

New research by Southwest Research Institute and the National Science Foundation's National Center for Atmospheric Research has produced a new tool that could eventually extend space weather forecasts from hours to weeks. The approach aims to provide earlier warnings of solar activity that can disrupt GPS, power grids, satellites and astronaut operations.

The work targets a longstanding heliophysics challenge: anticipating where and when large, flare-producing active regions will emerge on the Sun. According to SwRI scientist Subhamoy Chatterjee, these solar active regions exhibit tangled magnetic fields and generate explosive events such as solar flares and coronal mass ejections, which can drive hazardous space weather.

Solar active regions tend to cluster along large-scale, warped magnetic toroidal bands rather than appearing randomly across the solar surface. Using magnetic measurements from the Solar Dynamics Observatory's Helioseismic and Magnetic Imager taken on February 14, 2024, the team showed that surface magnetic patterns can be inverted to reconstruct critical subsurface magnetic states.

Most operational space weather tools currently rely on localized magnetic signatures that only become clearly predictive a few hours before an eruption. To push forecasts further ahead, the SwRI and NSF-NCAR team developed PINNBARDS, a Physics-Informed Neural Network-Based Active Region Distribution Simulator, which links visible solar active regions to the deep magnetic dynamics of the Sun's tachocline.

The tachocline is a thin transition layer between the uniformly rotating radiative interior and the more turbulent outer convection zone, and it plays a key role in solar magnetic field generation. By incorporating physical constraints into the neural network, PINNBARDS uses global surface magnetic information to infer how large-scale magnetic structures evolve in this deep region.

By bridging surface observations and deep solar magnetic processes, the researchers are advancing a physics-informed, AI-enabled framework for anticipating extreme space weather events. The global perspective built into PINNBARDS offers the potential for significantly longer forecast lead times, a capability viewed as critical for protecting communications networks, satellites and future human spaceflight.

"The reconstructed subsurface states from PINNBARDS provide initial conditions for forward simulations of solar magnetic evolution, opening the door to predicting where and when large, flare-producing active regions are likely to emerge weeks in advance," said Mausumi Dikpati, a senior scientist at NSF-NCAR who led the team. The latitude and longitude at which new active regions appear are especially important because they determine whether resulting solar particle bursts are likely to intersect Earth's orbit.

Research Report:A Physics Informed Neural Network for Deriving MHD State Vectors from Global Active Regions Observations