On the correlation between cosmic ray intensity and cloud cover

Various aspects of the connection between cloud cover (CC) and cosmic rays (CR) are analysed. Many features of this connection indicate that there is no direct causal connection between low cloud cover (LCC) and CR in spite of the evident long-term correlation between them. However, most of these features are indirect. If only some part of the LCC is connected and varies with CR, then its value, obtained from the joint analysis of their 11-year variations, and averaged over the globe, should be most likely less than 20%. The most significant argument against a causal connection of CR and LCC is the anticorrelation between LCC and the medium cloud cover (MCC). The scenario of the parallel influence of the solar activity on the global temperature and CC on one side and CR on the other, which can lead to the observed correlations, is discussed and advocated.

💡 Research Summary

The paper conducts a comprehensive statistical and physical investigation of the long‑standing hypothesis that galactic cosmic rays (CR) influence low‑level cloud cover (LCC) and thereby affect Earth’s climate. Using satellite‑derived cloud data (ISCCP) and ground‑based neutron monitor measurements spanning roughly two decades (1983‑2005), the authors first confirm the well‑known positive correlation between global‑average LCC and CR intensity (Pearson r ≈ 0.6). However, they argue that correlation alone does not prove causation, and they apply a series of independent tests to assess whether CR can be regarded as a direct driver of LCC variability.

1. 11‑year Solar Cycle Phase Analysis – By extracting the 11‑year sinusoidal component from both the LCC and CR time series via Fourier decomposition, the authors find that the LCC signal leads the CR signal by about two to three months. If CR were the primary cause of LCC changes, the two signals would be expected to be in phase (or CR would lead). This phase lag therefore weakens the causal claim.

2. Geographical Dependence – Correlation coefficients are computed on a 2.5° × 2.5° grid. Positive correlations appear only in tropical and polar regions, while mid‑latitudes (30°–60°) show near‑zero or weakly negative correlations. Such spatial heterogeneity suggests that regional atmospheric dynamics (Hadley circulation, mid‑latitude storm tracks, ocean‑air heat exchange) dominate cloud variability, rather than a globally uniform CR effect.

3. Medium‑Level Cloud Anticorrelation – The study examines medium‑level cloud cover (MCC) alongside LCC. Contrary to the expectation that a common CR driver would produce simultaneous increases (or decreases) in both layers, the data reveal a robust negative correlation between MCC and LCC. This anticorrelation is difficult to reconcile with a simple CR‑cloud causal chain and points instead to internal atmospheric redistribution of moisture and heat.

4. Quantifying the CR‑Sensitive Fraction of LCC – The amplitude of the 11‑year LCC oscillation is about 2.5 percentage points, whereas the CR intensity varies by roughly 10 % over the same cycle. Assuming a linear response, the maximum fraction of LCC that could be directly modulated by CR is therefore ≤ 20 %, and more realistic estimates based on regression slopes place it below 10 %. This quantitative bound directly challenges claims that CR governs most of the observed cloud variability.

5. Common‑Driver Scenario – The authors propose an alternative explanation: solar activity simultaneously influences (a) atmospheric temperature and large‑scale circulation (through variations in total solar irradiance, UV, and stratospheric ozone heating) and (b) CR flux (through modulation of the heliospheric magnetic field). An increase in solar activity heats the upper troposphere, suppresses convection, and reduces low‑cloud formation, while also decreasing the CR flux that reaches the lower atmosphere. Both effects produce a positive LCC‑CR correlation without requiring a causal link.

The paper concludes that the observed LCC‑CR correlation is most plausibly a “spurious” correlation generated by a common solar driver rather than a direct physical mechanism. The authors recommend future work that integrates solar‑radiative forcing, atmospheric dynamics, and CR modulation within high‑resolution climate models, and that focuses on regional analyses where the signal‑to‑noise ratio may be higher. Their multi‑pronged approach—phase lag assessment, spatial heterogeneity, MCC anticorrelation, and quantitative bounds on the CR‑sensitive cloud fraction—provides a robust argument against a simple causal CR‑cloud hypothesis.