Geomagnetic storms, disturbances in Earth’s magnetic field caused by solar activity, can disrupt vital systems like power grids, GPS, and satellite communications. Most current forecasts focus on global averages, which miss important differences from one region to another. This means that some areas may face stronger effects than predicted.

To improve local forecasts, we developed a machine learning model that predicts the Local Disturbance Index (LDi), which measures geomagnetic activity at specific locations. Our model uses real-time space weather data from the ACE satellite, along with details about each observatory’s position and local time, to predict disturbances 1 to 2 hours in advance. It also includes confidence intervals to help assess prediction reliability.

To test how well the model generalizes, we trained it on data from five observatories and then tested it on a sixth, previously unseen location. We evaluated the model by comparing its predictions to the natural difference between local and global values. The results show that our model consistently outperforms this baseline, providing more accurate and location-specific forecasts.

This work shows that localized forecasting is not only possible but essential for improving space weather resilience in specific regions.