Correspondences
What is linked, what is not
Each arc joins two phenomena tested on this site. Green and animated: clear link (correlation significant at 5 % after correcting for autocorrelation, or a response outside chance in an event study). Grey: inconclusive test. A statistical link alone does not establish causation.
- Clear link
- Inconclusive
Hypotheses tested
32
Clear links
23
Causation demonstrated
0
Arc width from |r| of the stored result; several tests between the same phenomena are drawn as stacked arcs.
Hypothesis tested: Galactic cosmic rays, more numerous when the Sun is quiet, ionise the air; the ions might help form condensation nuclei and hence low clouds, which would link the solar cycle to climate (Svensmark and Friis-Christensen, 1997; Marsh and Svensmark, 2000). The correlation found with ISCCP low clouds over 1983–1994 faded afterwards (Sloan and Wolfendale, 2008; Laken et al., 2012), and the CLOUD experiment at CERN found ion-induced nucleation too weak to matter for today’s clouds (Dunne et al., 2016). Expected here: no robust link.
Pearson correlation r (lag 0 month(s))
0.16
95% interval [-0.18; 0.43] (block bootstrap)
Aligned points
408
effective n ≈ 68 (autocorrelation)
p-value
0.186
two-sided, Fisher z transform, effective n
r from 1983 to 1994
0.59
p = 0.004, n = 138
r from 1995 to 2017
-0.29
p = 0.070, n = 270
Period analysed
1983-07 → 2017-06
computed on Oct 26, 2026 · engine v1.0
Aligned series (standard deviations, after preprocessing)
Show the data as a table
Correlation by lag (month(s))
Show the data as a table
By cloud level and period
| Clouds (ISCCP-H) | 1983–2017, detrended | 1983–2017, raw | 1983–1994 | 1995–2017 |
|---|---|---|---|---|
| Low (IR) | 0.16 (p = 0.19) | -0.15 | 0.59 (p = 0.00) | -0.29 (p = 0.07) |
| Middle (IR) | -0.06 (p = 0.66) | -0.17 | -0.36 (p = 0.30) | 0.32 (p = 0.00) |
| High (IR) | 0.03 (p = 0.84) | -0.20 | 0.12 (p = 0.65) | 0.16 (p = 0.34) |
| Total cloud amount | 0.07 (p = 0.59) | -0.33 | 0.24 (p = 0.24) | 0.00 (p = 0.98) |
Method and sources
Cosmic-ray flux: monthly mean of the daily counts of the Oulu neutron monitor (months 80 % covered), as a % of the mean. Clouds: global means weighted by cos(latitude) of the ISCCP-H monthly files (NOAA NCEI, 1983–2017), low, middle and high infrared cloud types and total cloud amount. Both series are turned into anomalies (mean of each calendar month removed), then their linear trend is removed: the ISCCP record contains spurious trends due to changes of satellites and viewing angles (Evan et al., 2007).
Lag fixed at 0 in advance (ion nucleation would act within days); curve from −12 to +12 months shown. Bretherton et al. effective size; block bootstrap with 24-month blocks. Same calculation, without detrending as in the original papers, over 1983–1994 (period of the claim) then 1995–2017, and for the other cloud levels.
Computation parameters
{
"lag": "fixed at 0 a priori; curve from −12 to +12 months",
"clouds": "ISCCP-H global means weighted by cos(latitude), IR cloud types (low, middle, high) and total",
"filter": "anomalies from the mean of each calendar month, then linear trend removed (whole period)",
"bootstrap": {
"block": 24,
"draws": 1000
},
"statistic": "pearson",
"subperiods": "to 1995-01 (period of the original claim) and after, anomalies without detrending",
"cosmic_rays": "Oulu neutron monitor, monthly mean of the daily values (months 80 % covered), % from the mean"
}Inputs : oulu.nm.daily (22,811 pts, run #3730) · isccp.cloud.low.monthly (408 pts, run #4304) · isccp.cloud.mid.monthly (408 pts, run #4304) · isccp.cloud.high.monthly (408 pts, run #4304) · isccp.cloud.total.monthly (408 pts, run #4304)
Reading this result
Over 1983–1994 the published correlation is indeed found (r ≈ 0.6 with low clouds). But it changes sign after 1995, does not exist over the whole 34 years once the trend is removed, and is not found from one cloud level to another (middle clouds respond the opposite way). The lag curve peaks when clouds lead cosmic rays, the sign of a shared slow variation (artefacts of the satellite record, solar cycle) rather than of a physical effect. Conclusion in line with the literature: no robust link between cosmic rays and clouds at the global scale. A negative test published as such. Correlation is not causation: common factors (seasons, cycles, trends) or coincidence can produce it. The p-values account for autocorrelation (effective n) but not for preprocessing choices; the confidence interval comes from a block bootstrap. The limits of the source data apply.
SOURCES OF THIS PAGE
- Oulu neutron monitor, daily corrected count rate (Sodankylä Geophysical Observatory)
-
Oulu neutron monitor data, cosmicrays.oulu.fi, Sodankylä Geophysical Observatory of the University of Oulu (6NM64 standard neutron monitor at 65.05° N, 25.47° E, geomagnetic cutoff rigidity ≈ 0.8 GV).
· source page ↗
Licence : Research, educational and public non-commercial use; acknowledge cosmicrays.oulu.fi and the Sodankylä Geophysical Observatory (commercial use: contact the station manager) · retrieved on Oct 25, 2026 · SHA-256 fc2f77ecf31e… - ISCCP-H cloud amount, monthly global means (NOAA NCEI)
-
Young, A. H., Knapp, K. R., Inamdar, A., Hankins, W., & Rossow, W. B. (2018). The International Satellite Cloud Climatology Project H-Series climate data record product. Earth System Science Data, 10, 583–593.
· doi:10.5194/essd-10-583-2018 · source page ↗
Licence : U.S. Government work, public domain; cite Young et al. (2018) · retrieved on Oct 26, 2026 · SHA-256 aaaa354f36d0… - Svensmark et al. (1997)
-
Svensmark, H., & Friis-Christensen, E. (1997). Variation of cosmic ray flux and global cloud coverage — a missing link in solar-climate relationships. Journal of Atmospheric and Solar-Terrestrial Physics, 59, 1225–1232.
· doi:10.1016/S1364-6826(97)00001-1
Licence : Published literature; cited - Marsh et al. (2000)
-
Marsh, N. D., & Svensmark, H. (2000). Low cloud properties influenced by cosmic rays. Physical Review Letters, 85, 5004–5007.
· doi:10.1103/PhysRevLett.85.5004
Licence : Published literature; cited - Sloan et al. (2008)
-
Sloan, T., & Wolfendale, A. W. (2008). Testing the proposed causal link between cosmic rays and cloud cover. Environmental Research Letters, 3, 024001.
· doi:10.1088/1748-9326/3/2/024001
Licence : Published literature; cited - Laken et al. (2012)
-
Laken, B. A., Pallé, E., Čalogović, J., & Dunne, E. M. (2012). A cosmic ray-climate link and cloud observations. Journal of Space Weather and Space Climate, 2, A18.
· doi:10.1051/swsc/2012018
Licence : Published literature; cited - Dunne (2016)
-
Dunne, E. M., et al. (2016). Global atmospheric particle formation from CERN CLOUD measurements. Science, 354, 1119–1124.
· doi:10.1126/science.aaf2649
Licence : Published literature; cited - Evan et al. (2007)
-
Evan, A. T., Heidinger, A. K., & Vimont, D. J. (2007). Arguments against a physical long-term trend in global ISCCP cloud amounts. Geophysical Research Letters, 34, L04701.
· doi:10.1029/2006GL028083
Licence : Published literature; cited