Cosmic Ray Inter-Station Correlation Variations as Precursors of Geomagnetic Storms: A Statistical Study and Multi-Parameter Early Warning Framework

The modulation of galactic cosmic rays (GCRs) by interplanetary disturbances, manifested as Forbush decreases (FDs), has long been recognized as a signature of coronal mass ejection (CME) passages through the heliosphere. While individual FD events have been extensively studied, systematic investigations of how GCR inter-station correlation variations relate to geomagnetic storm (GS) intensity have not been established. Here we analyze the relationship between GCR characteristics (from a global NM network) and GSs, aiming to understand the physical mechanisms of heliospheric disturbances and to develop complementary predictive capabilities beyond existing L1 solar wind monitoring. By applying a newly introduced anisotropy characteristic method alongside correlation analysis to 25 years of hourly NM data (1995-2020, seven stations), we demonstrate significant correlations between GCR parameters and geomagnetic activity. Inter-station relative differences and anisotropy enhancements show distinct precursor signatures depending on storm intensity, with extreme events displaying detectable signals 48-96 hours in advance. Based on these intensity-dependent response patterns, we propose a “two-stage multi-level” early warning framework: mid-term identification (48-96 hr) triggered by sustained anisotropy increases, followed by short-term grading (0-48 hr) based on inter-station relative difference variations and high-latitude flux changes. Validated on the extreme November 2003 and severe August 2018 geomagnetic storms, our approach successfully identifies precursor signals, providing a potential means to extend GS prediction windows.

💡 Research Summary

**
The paper investigates whether variations in galactic cosmic‑ray (GCR) measurements can serve as reliable precursors to geomagnetic storms (GS) and, if so, how they can be incorporated into an operational early‑warning system that extends beyond the limited lead times offered by L1 solar‑wind observations. Using a 25‑year (1995‑2020) dataset of hourly pressure‑corrected count rates from seven globally distributed neutron‑monitor (NM) stations (OULU, JUNG, SOPO, THUL, KERG, PTFM, TSMB) together with the hourly Dst index, the authors analyse 280 808 hours covering 2 809 storm events across four intensity categories (minor, moderate, severe, extreme).

Two novel GCR‑derived parameters are introduced. The first, the inter‑station relative difference (δ), is defined as the normalized difference between the high‑latitude OULU and mid‑latitude JUNG stations, effectively removing common‑mode variations (e.g., atmospheric pressure) and isolating north‑south asymmetries caused by CME magnetic structures. The second, a simplified anisotropy metric (A_basic), is the sum of squared deviations of each station’s count rate from its quiet‑time baseline (A_basic(t)=∑ R_i(t)^2). Unlike traditional spherical‑harmonic anisotropy analyses, A_basic can be computed in real time from raw NM data.

Correlation analyses reveal that both δ and A_basic are significantly linked to Dst (p < 0.001) and that the strength of the correlation increases with storm intensity. For minor storms, GCR flux changes lead Dst by 9–17 h, providing a modest early‑warning window. For severe and extreme storms, the peak correlations occur within 0–20 h of storm onset, indicating near‑simultaneous evolution of GCR signatures and geomagnetic response. Notably, A_basic shows a pronounced rise 48–96 h before extreme events, while δ reaches its maximum correlation in the 0–24 h window, especially for severe storms (r_s ≈ −0.57). High‑latitude stations (OULU, THUL, SOPO) consistently exhibit stronger correlations than low‑latitude sites, reflecting their lower geomagnetic cut‑off rigidities.

Based on these findings, the authors propose a “two‑stage multi‑level” early‑warning framework. Stage 1 (mid‑term, 48–96 h) triggers an alert when A_basic exceeds a statistically defined threshold, indicating a sustained anisotropy enhancement. Stage 2 (short‑term, 0–48 h) refines the warning by monitoring δ and rapid flux drops at high‑latitude stations, allowing a graded severity rating. The framework is validated against two benchmark storms: the extreme November 2003 “Halloween” storm (Dst ≈ −422 nT) and the severe August 2018 storm (Dst ≈ −210 nT). In both cases, the anisotropy metric signaled the impending event 60–72 h in advance, while the relative‑difference metric provided a high‑confidence short‑term alert within 12 h of onset. This lead time substantially exceeds the 30–60 min window provided by L1 solar‑wind monitors such as ACE or DSCOVR.

The study concludes that GCR‑based parameters, especially a real‑time anisotropy index and inter‑station differential, are robust, intensity‑dependent precursors to geomagnetic storms. Their integration into operational space‑weather pipelines could extend warning horizons, reduce false alarms associated with CME propagation uncertainties, and improve preparedness for extreme space‑weather events. Future work is suggested to incorporate additional NM stations, explore machine‑learning classification of precursor patterns, and combine GCR metrics with traditional solar‑wind and coronagraph observations for a comprehensive probabilistic forecasting system.