Towards a fully automated end-to-end pipeline for short-term CME magnetic field forecasting

We present a fully automated end-to-end pipeline for operational short-term forecasting of the in situ magnetic structure of coronal mass ejections (CMEs) at Earth. Triggered by new events in the NASA/CCMC DONKI catalog, the system couples ensemble arrival time predictions using ELEvo with deep-learning-based in situ detection of magnetic obstacles (ARCANE), and iterative flux rope reconstruction using the semi-empirical 3DCORE model. As more real-time L1 solar wind data becomes available, the pipeline continuously updates forecasts of the remaining CME magnetic field profile. Using archived real-time data, we evaluate the pipeline under operational constraints and analyze how reconstruction quality evolves as a function of available data, providing insight into capabilities and limitations of fully automated real-time CME magnetic field reconstruction for space weather forecasting.