Verbesserte Sonnensturmmodellierung mit Machine Learning

Solar storms (CMEs) can cause geomagnetic disturbances that affect technology on Earth. Wide-angle imaging enables their tracking but is often incomplete in real time. We use machine learning to enhance these data and develop a CME tracking tool – a contribution to ESA’s Vigil mission and improved space weather forecasting.

Solar storms, also known as coronal mass ejections (CMEs), are large eruptions of plasma and magnetic fields from the solar corona. During solar maximum, roughly one CME per week reaches Earth, potentially triggering geomagnetic storms. These storms can disrupt power grids, satellites, and communication systems. In extreme cases, they may damage transformers and cause widespread power outages.

To forecast the arrival of CMEs, observations are required to model the storm’s path to Earth. Coronagraphs provide real-time data but only up to about 30 solar radii—roughly ten percent of the Sun-Earth distance. In contrast, Heliospheric Imagers (HI) cover the entire space between the Sun and Earth, allowing CMEs to be tracked along their full trajectory. These data are ideal for modeling CME kinematics and arrival times but are often low-resolution and incomplete in real time. High-quality HI data become available only days later and are thus not suitable for real-time forecasting.

In this project, we aim to combine HI observations with machine learning methods to improve CME arrival predictions. We pursue two approaches:

Enhancing real-time HI data: We train an algorithm using both high- and low-quality HI data to generate artificially improved real-time data. These enhanced data are then used to improve the accuracy of our prediction model.

Automated CME detection and tracking: We are developing a tool to automatically detect and track CMEs in HI data. Such tools currently exist only for coronagraphs, where Earth-directed CMEs are often difficult to identify.

Both approaches aim to improve forecast accuracy and reduce false alarms. In the context of ESA’s Vigil mission, this project makes an important contribution to space weather forecasting based on HI data.