Abstract Space hurricanes are distinct space weather phenomena that occur during extremely quiet geomagnetic conditions, and exhibit hurricane‐like cyclonic auroral bright spot structures. This phenomenon can induce severe space weather effects, including radio communication disruptions, navigation and positioning errors, and over‐the‐horizon radar detection impairments. Previous identification relied mainly on manual inspection of space hurricanes, which is inefficient, subjective, and lacks tools for high‐precision automatic recognition. To address these challenges, we developed a deep learning model by incorporating attention mechanisms and multi‐scale feature extraction, trained on more than a decade of satellite‐based extreme ultraviolet (EUV) image data. Through systematic hyperparameter optimization and adaptive learning rate scheduling, the model achieves high‐precision automatic identification and pixel‐level localization of space hurricanes, reaching an accuracy of 97.90% on a challenging global data set. We also developed an end‐to‐end detection and localization system with visual interactive capabilities. This research provides crucial support for modeling of polar space weather and lays a foundation for developing space environment risk warning and adaptive regulation systems.
Li et al. (Fri,) studied this question.