Speaker
Description
Accurate and timely estimates of CME direction, speed and size are essential for space weather forecasting; however, current methods based on coronagraph observations remain time-consuming and show significant variability between forecasters. To address this challenge, we propose an automated system capable of continuously producing a 3D reconstruction of CMEs and inferring their kinematics as a step forward for operational forecasting. Our new method, based on a physical 3D forward model similar to the Graduated Cylindrical Shell (GCS), allows us to automatically fit the model to multi-viewpoint coronagraph images. Using gradient-based backpropagation, the model's parameters (e.g. direction, tilt, height) are optimised automatically by direct comparison to the CME's outline. Furthermore, by incorporating uncertainty estimates into the CME masks, we can propagate these uncertainties through the forward-model fitting, providing confidence intervals on the reconstructed parameters, improving reproducibility but also enhancing the physical interpretability of the results compared to purely data-driven CNN approaches.
The method was evaluated on several CME events observed by STEREO COR2 and LASCO C3, with performance assessed both visually (by comparing reconstructed CME fronts with observed outlines across multiple viewpoints) and quantitatively, by measuring consistency in derived direction, angular width, and propagation speed. We further compared our results to previous GCS-based manual reconstructions, showing that the automatic approach achieves comparable accuracy.
These results highlight the potential of the method for integration into operational pipelines, where its ability to deliver fast and reproducible reconstructions makes it well-suited for real-time space weather monitoring and forecasting. Finally, by focusing on STEREO positions around the Lagrangian points 4 and 5, these results lay the groundwork for preparing for ESA’s Vigil mission and investigating synergies with future L1 missions, including those following PUNCH.
