The Frequency‐Domain Characterization of Cosmic Ray Intensity Variations Before Forbush Decreases Associated With Geomagnetic Storms

Abstract

Nonrecurrent geomagnetic storms caused by Coronal Mass Ejection (CME) can induce serious impacts on space‐ and ground‐based equipment. However, these nonrecurrent geomagnetic storms are hard to predict since CMEs are not periodic. Previous studies have shown that the variations of Cosmic Ray Intensity (CRI) before nonrecurrent storms may forebode the coming geomagnetic storm. But it is difficult to extract the variations since the cosmic ray flux is a complex signal. In order to identify the precursory signal in CRI variations triggered by CME, an ensemble self‐adaptive time‐frequency analysis method was proposed and applied to 65 nonrecurrent geomagnetic storms occurred between 1998 and 2019. The results indicate that the precursory signals are successfully identified in 43 of 45 storms, after excluding 20 storms accompanied by ground level enhancement. In addition, it is also found that the spectral density of the precursory signal could reflect the active level of geomagnetic condition, which is lower during geomagnetic quiet period.

A New Method for Predicting Non‐Recurrent Geomagnetic Storms

“…We regarded the time between the precursor amplitude reaching the threshold to the arrival of a shock to the earth as the leading time of this model. According to the threshold value of 1.2 times the base value, the overall average leading time can reach 50.4 hr, which is better than the result supported by Ye et al (2022). Moreover, with the decrease of geomagnetic storm strength, the leading time will gradually increase because the speed of CME that can produce strong geomagnetic storms is usually faster, and the time of shock generated by them to reach the Earth is shorter (Gopalswamy et al, 2007).…”

“…Though the variation of CRI in the time domain before storms, the occurrence could be predicted more than 6-12 hr in advance, even 20 hr in advance (Dorman, 2005;Lingri et al, 2019;Munakata et al, 2000;Papailiou et al, 2012). Similarly, by analyzing the variation of CRI in the frequency domain, the precursor of the geomagnetic storm can be extracted to predict the occurrence of geomagnetic storms with lead times of about 20 hr (Adhikari et al, 2019;Kudela et al, 2001;Ye et al, 2022;Zhu et al, 2015). Ye et al (2022) used the time-frequency analysis CEEMDAN-CWT (Complete Ensemble Empirical Mode Decomposition with Adaptive Noise-Continuous Wavelet Transform) method to investigate 65 non-recurrent geomagnetic storms (Kp ≥ 7) that occurred during the 23rd and 24th solar activity cycle, and pointed out that geomagnetic storm precursor signals reconstructed from CRI by using this method could predict the occurrence of non-recurrent geomagnetic storms 22.5 hr in advance.…”

“…Similarly, by analyzing the variation of CRI in the frequency domain, the precursor of the geomagnetic storm can be extracted to predict the occurrence of geomagnetic storms with lead times of about 20 hr (Adhikari et al, 2019;Kudela et al, 2001;Ye et al, 2022;Zhu et al, 2015). Ye et al (2022) used the time-frequency analysis CEEMDAN-CWT (Complete Ensemble Empirical Mode Decomposition with Adaptive Noise-Continuous Wavelet Transform) method to investigate 65 non-recurrent geomagnetic storms (Kp ≥ 7) that occurred during the 23rd and 24th solar activity cycle, and pointed out that geomagnetic storm precursor signals reconstructed from CRI by using this method could predict the occurrence of non-recurrent geomagnetic storms 22.5 hr in advance. Despite its effectiveness, the complexity of CRI limits its ability to remove the influence of ground level enhancement (GLE) events and predict moderate and small storms (Kp ≤ 6) (Ye et al, 2022).…”

“…The periodic signals in CRI are due to the fact that the cosmic ray is modulated by the heliosphere magnetic field, solar wind, and Earth magnetosphere (Jeong & Oh, 2022;Kravtsova & Sdobnov, 2016;Mandrikova et al, 2018), while most of the non-recurrent signals in CRI are due to the occasional activity of the sun, such as CME, flare. According to Ye et al (2022), there are precursor signals associated with CME-driven geomagnetic storms could be identified in the frequency domain of CRI using the time-frequency analysis method CEEMDAN-CWT. However, the weaker precursor signals associated with CME-driven geomagnetic storms will be obscured by periodic changes, making this method only suitable for stronger storms (Kp ≥ 7).…”

“…Ye et al (2022) used the time-frequency analysis CEEMDAN-CWT (Complete Ensemble Empirical Mode Decomposition with Adaptive Noise-Continuous Wavelet Transform) method to investigate 65 non-recurrent geomagnetic storms (Kp ≥ 7) that occurred during the 23rd and 24th solar activity cycle, and pointed out that geomagnetic storm precursor signals reconstructed from CRI by using this method could predict the occurrence of non-recurrent geomagnetic storms 22.5 hr in advance. Despite its effectiveness, the complexity of CRI limits its ability to remove the influence of ground level enhancement (GLE) events and predict moderate and small storms (Kp ≤ 6) (Ye et al, 2022). CRI exhibits a complex structure that includes both non-recurrent signals, such as GLE, FD, and CME-driven fluctuations (Belov & Gushchina, 2018;Dorman, 1999) and periodic signals, such as 22-year cycle, 11-year cycle, annual variation, seasonal variation, diurnal variation, and so on (Chowdhury et al, 2016;Jeong & Oh, 2022;Krymsky et al, 2005;Shen & Qin, 2017;Tezari et al, 2016).…”