A New Method for Predicting Non-Recurrent Geomagnetic Storms

Abstract

Predicting non-recurrent geomagnetic storms caused by Coronal Mass Ejection (CME) is important and necessary for space weather forecasting. Previous studies have shown that it is feasible to predict non-recurrent geomagnetic storms by reconstructing precursors from cosmic ray intensity (CRI) in the frequency domain. However, the difficulty lies in predicting the minor storms, the moderate storms, and the storms accompanying ground level enhancement (GLE). This study proposes a new method that includes the spectral whitening method and CEEMDAN-CWT for predicting non-recurrent geomagnetic storms. The method was applied to 229 CME-driven events, including 166 events with Kp ≥ 5 and 63 events with Kp < 5, during the solar cycles 23 and 24. This study analyzed 166 geomagnetic storm events and found that 129 of them were accurately predicted, resulting in a recall rate of 77.7%. Additionally, the study found that as the Kp index increased, the amplitude of the precursor detected by this method also increased, while the time interval between the onset of CME and the maximum amplitude of the precursor decreased.