AI could have predicted the whole chain of events associated with the may 2024 space weather superstorm

This talk applies artificial intelligence (AI) to predict the whole chain of events associated with the May 2024 superstorm, including solar flares from NOAA active region (AR) 13644, Earth-directed CMEs, and a violent geomagnetic storm. Specifically, we will show that, using magnetogram cut-outs, a Vision Transformer was able to classify the evolution of the AR morphology and a video-based deep learning could have predicted the occurrence of solar flares; further, using remote sensing and in-situ observations, we will show that physics-driven models were able to improve the accuracy of CME travel time prediction and provide timely alert of its geomagnetic impact.  The results showed unprecedented accuracy in predicting both solar flares and geomagnetic disturbance occurrences, as well as CME arrival, with uncertainty as low as one minute.