RealClimate logo

Cosmic rays and clouds: Potential mechanisms

Filed under: — group @ 26 September 2011

Guest Commentary by Jeffrey Pierce (Dalhousie U.)

I’ve written this post to help readers understand potential physical mechanisms behind cosmic-ray/cloud connections. But first I briefly want to explain my motivation.

Prior to the publication of the aerosol nucleation results from the CLOUD experiment at CERN in Nature several weeks ago Kirkby et al, 2011, I was asked by Nature Geoscience to write a “News and Views” on the CLOUD results for a general science audience. As an aerosol scientist, I found the results showing the detailed measurements of the influences of ammonia, organics and ions from galactic cosmic rays on aerosol formation exciting. While none of the results were entirely unexpected, the paper still represents a major step forward in our understanding of particle formation. This excitement is what I tried to convey to the general scientific audience in the News and Views piece. However, I only used a small portion of the editorial to discuss the implications to cosmic rays and clouds because (1) I felt that these implications represented only a small portion of the CLOUD findings, and (2) the CLOUD results address only one of several necessary conditions for cosmic rays to affect clouds, and have not yet tested the others.

Many of the news articles and blog posts covering the CLOUD article understandably focused much more on the cosmic-ray/cloud connection as it is easy to tie this connection into the climate debate. While many of the articles did a good job at reporting the CLOUD results within the big picture of cosmic-ray/cloud connections, some articles erroneously claimed that the CLOUD results proved the physics behind a strong cosmic-ray/cloud/climate connection, and others still just got it very muddled. A person hoping to learn more about cosmic rays and clouds likely ended up confused after reading the range of articles published. This potential confusion (along with many great questions and comments in Gavin’s CLOUD post) motivated me to write a general overview of the potential physical mechanisms for cosmic rays affecting clouds. In this post, I will focus on what we know and don’t know regarding the two major proposed physical mechanisms connecting cosmic rays to clouds and climate.

What we know and don’t know about the connection between cosmic rays and clouds and climate

These two proposed mechanisms are the ion-aerosol clear-sky hypothesis and the ion-aerosol near-cloud hypothesis (using the terminology from Carslaw et al., 2002). The ion-aerosol clear-sky hypothesis has gotten most of the mainstream attention, and the recent CLOUD results test a portion of this hypothesis. The near-cloud hypothesis has received less attention. I believe this is because little is known about many of the processes involved. Regardless, it is a fascinating and plausible hypothesis, so I will also address it here. The central question we need to answer in either of these hypotheses is “How much do clouds change due to a change in cosmic rays?”.

The ion-aerosol clear-sky hypothesis

The central theme of the clear-sky hypothesis is that cosmic rays affect ion concentrations in the atmosphere. Aerosol nucleation (the formation of ~1 nm particles in the atmosphere) is generally enhanced by the presence of ions. The particles formed through nucleation may grow through condensation of sulfuric acid and organic vapors to sizes where they can act as Cloud Condensation Nuclei (CCN) (the particles on which cloud drops form). If CCN are exposed to relative humidities above 100%, cloud droplets will form on them. Thus, a change in cosmic rays could potentially affect the number of cloud drops, which in turn may affect the amount of sunlight reflected by a cloud, the formation of precipitation and the cloud lifetime.

Figure 1. Overview of freshly nucleated particles, CCN and cloud droplets.

For us to understand the clear-sky hypothesis, and answer the question, “How much do clouds change due to a change in cosmic rays?”, we must understand the following sub-questions:

  1. How much does ion formation in the atmosphere change due to changes in the cosmic-ray flux to the atmosphere (due to the solar cycle etc.)?
  2. How much do aerosol nucleation rates change due to changes in ion formation rates?
  3. How much do CCN concentrations change due to changes in aerosol nucleation rates?
  4. How much do clouds change due to changes in CCN concentrations?

Question 1: How much does ion formation in the atmosphere change due to changes in the cosmic-ray flux to the atmosphere?

Of the four questions, we understand question 1 the best. With current information about the Earth’s magnetic field and solar activity, we have fairly robust predictions of the ion formation rate from cosmic rays. The figure below shows the percent change in the ion formation rate from cosmic rays between the solar minimum (more cosmic rays) and solar maximum (fewer cosmic rays) Usoskin and Kovaltsov, 2006.

Figure 2. Percent change in the ion formation rate as a function of height and latitude in the atmosphere from cosmic rays between a typical solar minimum and solar maximum in the troposphere and lower stratosphere.

As shown in the figure above, the ion formation rate from cosmic rays varies by 5-20% throughout most of the troposphere (the region of the atmosphere where clouds form). The reported observed relative change in low cloud cover [4] is ~6% with the solar cycle (or 2% absolute change in the fraction that low clouds cover the planet). Thus, the modulation of ions is a similar order of magnitude to the amount of cloud change. In order for the clear-sky hypothesis to have a large effect on clouds, the 5-20% change in ion formation rates needs to efficiently propagate into changes in aerosol nucleation, CCN and cloud properties. So…

Question 2: How much do aerosol nucleation rates change due to changes in ion formation rates?

The recent CLOUD results in Nature directly address this question (and this question only). The results showed under the conditions of the CLOUD chamber show that ions from cosmic rays unequivocally aid aerosol nucleation. However, the CLOUD paper does not directly address how much nucleation rates will change from a 5-20% change in ion formation rates, but inspection of Figure 2 in their paper (below as our Figure 3) shows that a doubling of ion concentration leads to somewhat less than a doubling in nucleation rate. Furthermore, a doubling of ion concentration requires more than a doubling in ion formation rates (due to an increased rate of positive and negative ions re-combining with each other to form neutral molecules when ion concentrations are higher). Therefore, a 5-20% change in ion formation rates from cosmic-ray changes will lead to less than a 5-20% change in nucleation rates. (The results in Figure 3 covers a very large range in ion concentrations, much larger than would ever be modulated by relevant changes in cosmic rays.)

Figure 3. Figure 2 from Kirkby et al. (2011) showing the nucleation rate as a function of ion concentration for two different conditions (the two colored lines).

Question #3: How much do CCN concentrations change due to changes in aerosol nucleation rates?

The impact of changing aerosol nucleation rates on CCN concentrations has recently been studied using several different models Spracklen et al, 2008Makkonen et al, 2009Wang and Penner, 2009Yu and Luo, 2009. In all cases, the change in CCN is smaller than the change in nucleation rates. Two other papers Pierce and Adams, 2009, Snow-Kropla et al., 2011 have specifically looked at this question in the context of cosmic-ray changes, and found that even though nucleation rates are changing by 1-5% throughout much of the troposphere, the changes in CCN are generally around 0.1-0.2% throughout much of the globe. The reason for this strong dampening is shown in the figure below.

Figure 4. Schematic showing the reasons for the small changes in CCN to changes in nucleation rates

Firstly, primary emissions contribute to CCN as well as nucleation, and the primary emissions are not affected by cosmic rays. Secondly, the likelihood that a freshly nucleated particle will grow to become a CCN depends on whether it can grow from condensation of sulfuric acid and organic vapors onto it before the particle coagulates with a larger particle (reducing the number of particles). If the nucleation rate is increased due to cosmic rays, there will be more particles competing for a fixed amount of condensible vapors, and each new particle will grow more slowly. Additionally, the coagulation loss of the particles will increase due to the increased number of particles and the slower growth (particles are lost through coagulation more quickly at smaller sizes).

Unfortunately, as far as I know this question has only been addressed using models. While we test the model for known uncertainties in model inputs, it is always a possibility that we are missing something. Fortunately, the growth of ultrafine particles to CCN sizes should be addressed in future experiments in the CLOUD chamber, so we should soon also have controlled experimental evidence to compare with model results.

Question #4: How much do clouds change due to changes in CCN concentrations?

Increased CCN concentrations lead to increased concentrations of cloud droplets. More cloud droplets will lead to increased reflection of sunlight from the cloud to space, and may under some circumstances lead to a reduction of precipitation and an increased lifetime of the cloud. How much these cloud properties depend on CCN concentrations is a major area of research in general. CCN concentrations have more than doubled in many polluted regions due to human-generated emissions, so we are working hard to understand how this has affected clouds. Given that CCN concentrations have changed so much from human influence, a change in CCN of less than 1% due to cosmic rays seems quite minor. Indeed, cloud reflectivity, precipitation and cloud lifetime will generally change by less than the change in CCN for most clouds (e.g. we know that cloud cover has not more than doubled due to human-generated emissions). Therefore, it is unlikely to generate a ~6% change in cloud cover (reported in observations of clouds with 11-year solar cycle and after Forbush decreases) from less than a 1% change in CCN.

Clear-sky hypothesis summary

In summary, the clear-sky hypothesis is driven by 5-20% changes in ion formation rates in the troposphere. These ion changes would need to drive changes in cloud cover by several percent to account for reported correlations. While uncertainties in processes remain, it appears unlikely to me (and most other scientists working on aerosol-cloud interactions who’ve shared their thoughts on this hypothesis with me) that this mechanism will be strong enough to greatly change clouds. I would not go so far to say that the case is closed on this mechanism, but if it is to be important there must be some amplification factor in one (or more) of the questions described above that we are currently unaware of. Thus, it will be exciting to see what the future CLOUD experiments (or other controlled experiments) show regarding questions #3 and #4.

Ion-aerosol near-cloud hypothesis

The ion-aerosol near-cloud hypothesis has received less attention than the clear-sky hypothesis; however, there is still active research being done on it. The near-cloud hypothesis has to do with the global electric circuit (see the figure below).

Figure 5. Schematic showing how cosmic rays modulate the global electric circuit and may affect the charging around clouds.

Thunderstorms create a charge separation with positive ions at the top of the cloud and a negative ions at the bottom (this negative charge gets discharged through lightning to the ground). The positive charge at the top of the cloud moves through the conductive upper atmosphere to the ionosphere giving the ionosphere a positive charge. The difference in charge between the ionosphere and the Earth’s surface drives an electric current from the ionosphere to the surface. The resistance of the atmosphere to current flow depends on the ion concentrations (more ions = less resistance). Thus, when more cosmic rays enter the atmosphere, electricity flows more quickly through the atmosphere.

Non-thunderstorm clouds, however, interrupt the electric current because gas-phase ion concentrations within clouds are very low making the clouds very resistive to electric current flow. Charge builds up on the top and bottom of the cloud much like charged plates in a capacitor. Cosmic rays may affect this charge build up through changing the resistance of current flow in the clear atmosphere; however, the strength of this effect is still not well known.

This may have an effect on the cloud properties by enhancing the collision rate between cloud droplets and between aerosols and cloud droplets. Often in clouds, liquid water drops will exist even when temperatures are well below 0ºC (freezing point of water). Collisions between the charged aerosols with these supercooled cloud droplets may enable the freezing of these droplets, which could lead to cloud invigoration due to the heat released from freezing or enhanced precipitation (clouds consisting of both liquid drops and ice crystals are more effective at generating precipitations than clouds containing only one phase drops/crystals). These effects, however, are all still very uncertain.

Figure 6. The enhancement of droplet freezing by collisions with charged aerosol is an essential component of the near-cloud mechanism, but is not well understood.

The uncertainties in the near-cloud mechanism far exceed those of the clear-sky mechanism (it is not even clear whether a change in the cosmic-ray flux would lead to more or less cloud cover through the near-cloud mechanism). However, it remains an interesting potential connection between cosmic rays and clouds that needs to be explored if we are to understand how cosmic rays may affect clouds.

Final thoughts

While reported observed correlations between cosmic rays and clouds are suggestive of effects of cosmic rays on clouds, cosmic rays rarely change without other inputs to the Earth system also changing (e.g. total solar irradiance or solar energetic particle events, both also driven by changes in the sun, but distinct from cosmic rays). Thus, we must understand the physical basis of how cosmic rays may affect clouds. However, it is clear that substantially more work needs to be done before we adequately understand these physical connections, and that no broad conclusions regarding the effect of cosmic rays on clouds and climate can (or should) be drawn from the first round of CLOUD results. Finally, there has been no significant trend in the cosmic ray flux over the 50 years, so while we cannot rule out cosmic-ray/cloud mechanisms being relevant for historical climate changes, they certainly have not been an important factor in recent climate change.


  1. J. Kirkby, J. Curtius, J. Almeida, E. Dunne, J. Duplissy, S. Ehrhart, A. Franchin, S. Gagné, L. Ickes, A. Kürten, A. Kupc, A. Metzger, F. Riccobono, L. Rondo, S. Schobesberger, G. Tsagkogeorgas, D. Wimmer, A. Amorim, F. Bianchi, M. Breitenlechner, A. David, J. Dommen, A. Downard, M. Ehn, R.C. Flagan, S. Haider, A. Hansel, D. Hauser, W. Jud, H. Junninen, F. Kreissl, A. Kvashin, A. Laaksonen, K. Lehtipalo, J. Lima, E.R. Lovejoy, V. Makhmutov, S. Mathot, J. Mikkilä, P. Minginette, S. Mogo, T. Nieminen, A. Onnela, P. Pereira, T. Petäjä, R. Schnitzhofer, J.H. Seinfeld, M. Sipilä, Y. Stozhkov, F. Stratmann, A. Tomé, J. Vanhanen, Y. Viisanen, A. Vrtala, P.E. Wagner, H. Walther, E. Weingartner, H. Wex, P.M. Winkler, K.S. Carslaw, D.R. Worsnop, U. Baltensperger, and M. Kulmala, "Role of sulphuric acid, ammonia and galactic cosmic rays in atmospheric aerosol nucleation", Nature, vol. 476, pp. 429-433, 2011.
  2. K.S. Carslaw, "Cosmic Rays, Clouds, and Climate", Science, vol. 298, pp. 1732-1737, 2002.
  3. I.G. Usoskin, and G.A. Kovaltsov, "Cosmic ray induced ionization in the atmosphere: Full modeling and practical applications", Journal of Geophysical Research, vol. 111, 2006.
  4. H. Svensmark, and E. Friis-Christensen, "Variation of cosmic ray flux and global cloud coverage—a missing link in solar-climate relationships", Journal of Atmospheric and Solar-Terrestrial Physics, vol. 59, pp. 1225-1232, 1997.
  5. D.V. Spracklen, K.S. Carslaw, M. Kulmala, V. Kerminen, S. Sihto, I. Riipinen, J. Merikanto, G.W. Mann, M.P. Chipperfield, A. Wiedensohler, W. Birmili, and H. Lihavainen, "Contribution of particle formation to global cloud condensation nuclei concentrations", Geophysical Research Letters, vol. 35, 2008.
  6. R. Makkonen, A. Asmi, H. Korhonen, H. Kokkola, S. Järvenoja, P. Räisänen, K.E.J. Lehtinen, A. Laaksonen, V. Kerminen, H. Järvinen, U. Lohmann, R. Bennartz, J. Feichter, and M. Kulmala, "Sensitivity of aerosol concentrations and cloud properties to nucleation and secondary organic distribution in ECHAM5-HAM global circulation model", Atmospheric Chemistry and Physics, vol. 9, pp. 1747-1766, 2009.
  7. M. Wang, and J.E. Penner, "Aerosol indirect forcing in a global model with particle nucleation", Atmospheric Chemistry and Physics, vol. 9, pp. 239-260, 2009.
  8. F. Yu, and G. Luo, "Simulation of particle size distribution with a global aerosol model: contribution of nucleation to aerosol and CCN number concentrations", Atmospheric Chemistry and Physics, vol. 9, pp. 7691-7710, 2009.
  9. J.R. Pierce, and P.J. Adams, "Efficiency of cloud condensation nuclei formation from ultrafine particles", Atmospheric Chemistry and Physics, vol. 7, pp. 1367-1379, 2007.
  10. E.J. Snow-Kropla, J.R. Pierce, D.M. Westervelt, and W. Trivitayanurak, "Cosmic rays, aerosol formation and cloud-condensation nuclei: sensitivities to model uncertainties", Atmospheric Chemistry and Physics, vol. 11, pp. 4001-4013, 2011.

86 Responses to “Cosmic rays and clouds: Potential mechanisms”

  1. 51
    Gordon Jenkins says:

    Addendum to #24, Norilsk.

    Norilsk is above Arctic Circle at 69 degrees, 20 minutes (and spare change).
    MMC Norilsk Nickel, in addition to producing SO2, HS and CO2 also provides catalysts for nucleation and chemical reactions in the form of platinum, palladium and copper nano particles.

    Norilsk, twice per year falls under the aurora borealis (auroral oval) near the equinoxes and between 11pm and 2am local time (opposite sunside) during the periods of magnetic reconnection (THEMIS Satellite).

    Is it therefore inconceivable that the pollutants from Norilsk are involved in amplified nucleation effects due to auroral inductively coupled inloading, given the heavy metal particles emitted may be charged inductively like any other conductor passing through a magnetic field?

    Does Norilsk have the potential to meet all the requirements of CERN CLOUD and all of the concerns discussed so far?

    Would it not be prudent to take atmospheric samples in the Norilsk area and downwind?

    S. Fuchs, T.Hahn, H.G. Lintz, “The oxidation of carbon monoxide by oxygen over platinum, palladium and rhodium catalysts from 10−10 to 1 bar”, Chemical engineering and processing, 1994, V 33(5), pp. 363-369

  2. 52
    JimCA says:

    As a somewhat confused layman, I have a few questions trying to get past the details to the “big picture”. My apologies if these seem too simplistic, but I’m really just trying to get some perspective on all this…

    First, I assume it is not controversial that there seems to be at least a plausible causal connection from the solar wind to cosmic rays reaching our atmosphere to cloud formation to warming/cooling at the Earth’s surface.

    From this, I assume that variations in the solar wind presumably may have some (still to be determined) effect on global warming via this mechanism.

    Q1: What are the hard upper and lower bounds limiting the fraction of global temperature changes over the past century that might be attributed to variations in cosmic radiation? In particular, are 0% and 100% both plausible, or can tighter bounds be strongly supported?

    Q2: Same question, but looking forward a century?

    Q3: Same question, but looking at the past several million years?

    Q4: Is there a consensus range which most climate scientists would expect? (E.g., even if 0% and 15% are both possible, would either be surprising to most researchers?)

    Q5: Did the recent CERN results refine those ranges at all?

    Q6: Are future CERN results likely to refine those ranges?

    Thanks in advance to anyone willing to stick their neck out here… (and a pox on Captcha)

  3. 53
    CM says:

    Dr David King #49 (I’ll hazard a guess you’re not the former UK government science counselor by that name),

    We can measure cosmic radiation from Earth as it hits us, we don’t have to estimate it from the solar cycle. So complexities in the solar-GCR relationship, real or imagined, are surely not very relevant to testing the GCR-cloud connection.

    But relevant or not, please oh please do quantify the “substantial lag” from heliopause to Earth for particles traveling near the speed of light—relative to the 11-year solar cycle…

  4. 54
    Jeff Pierce says:

    #50 Gordon: This is an interesting idea. However, there are some complicating factors that make backing out cosmic ray effects from a single region (or a single plume in this case) difficult. (1) Meteorological conditions are always change and can greatly affect nucleation and growth. Thus, you’d need to focus on long-term measurements (e.g. 11-year solar cycle rather than Forbush decreases). (2) Emissions are never constant depending on the amount and composition of the ores they are smelting. My guess would be that the emissions will have much higher fractional variability than the changes in cosmic rays. Thus, measuring the changes in the 11-year cycle may have trends in emissions that may need to be corrected for, and this is not always easy. Thanks for the idea though!

  5. 55
    Jeff Pierce says:

    Interesting timing.

    It looks like we might be in the middle of a Forbush decrease right now…

    (note, this updates real time, so you view this in the future, it won’t show the decrease in Cosmic rays near the poles that I see.)

    There was a coronal mass ejection that just hit the Earth (

    muoncounter, you seem to know your stuff in this area. Can you weigh in on this?

  6. 56
    Jeff Pierce says:

    #51 JimCA: I’ll stick my kneck out a bit, but in no way would I consider myself an authority on most of the questions you are asking (Other than maybe question 5).

    Q1: Lower and upper bounds on the contribution of cosmic rays warming of past century?

    Lower bound: Essentially 0% of the warming of the past century was caused by cosmic rays. This would require us to better understand the reported correlations between cosmic rays and clouds and find that the correlations were not caused by the changes in cosmic rays (either the correlation was co-incidence or some other factor that correlates with cosmic rays [e.g. solar irradiance or solar energetic particle events] contributed to the changes in clouds). Furthermore, we’d need to show that the proposed physical mechanisms are weak.

    Upper bound: The cosmic ray flux decreased by about 5% between 1900-1950 (Carslaw et al., Science, 2002), but has not significantly changed since then. Thus, even in a case where cosmic rays do strongly affect clouds, they could only have greatly contributed to the early-20th century warming (~1/3 of the warming of the past century). The absolute upper bound would be a bit lower than this since GHGs were increasing during this period (albeit more slowly than recently). Thus, my best guess upper bound would be 25-30%. This would require us to (1) show in the mechanisms that cosmic rays very effectively change clouds, (2) show why cosmic rays don’t always correlate with cloud cover in observations, and (3) why the effect of changes in cosmic rays on clouds were so much bigger than the changes in human-generated aerosols (which increased very quickly in the first half of the 20th century) on clouds. This last point has always irked me a bit… we REALLY gunked up the atmosphere with particles.

    Q2 and Q3: The next century and past climates. I’m not an expert on the predictions of cosmic rays in the future or retrodictions of cosmic rays in the past. But it seems plausible that cosmic rays may have been a player in past climate changes. I really don’t have much knowledge here though.

    Q4: Consensus range for scientists. Speaking for others is the easiest way to get into trouble. Probably better to have an expert elicitation. I will say that most (though not all) aerosol and cloud physicists that I’ve spoken with think that the clear-sky mechanism (described in the post) is too weak to account for important changes in clouds. This research is my main area of focus, so I’m a bit more comfortable speaking on this.

    Q5: Do the CERN results refine those ranges at all. No. Certainly not the upper and lower bounds. As I said in the post, they have only touched on one part of clear-sky mechanism, and their results regarding cosmic rays are similar to previous results (their real advancements came by showing the chemical species in the nucleating cluster).

    Q5: Are the CERN results likely to refine those ranges? I hope so. I know this is their plan. We’ll see what they do. :)

  7. 57

    re: 49

    Thanks for the clarification. I believe my understanding on this has always been influenced by getting sun-burned on cloudy days when I was a kid.

  8. 58
    Jeff Pierce says:

    #20 Charles: Sorry for the slow response. I wrote one earlier, but it doesn’t appear to have posted properly. In my statement in this post, I was saying that 5-20% changes in cosmic rays appear unlike to change cloud cover by several percent through the clear-sky mechanism. In the Nat Geo article I was stating that ions from cosmic rays do make it more favorable for 1 nm particles to form. This does mean that there will necessarily be large changes in clouds from changes in cosmic rays.

    Thanks for the inline comment, Eric!

  9. 59
    wush says:

    Is it just the ionisation process which needs to be considered or can GCR’s have another effect?
    Quotes from above link:-
    While looking for climatic factors that might influence the growth of the trees, they made the surprising discovery that the trees grew faster in a pattern that matched with cycles of galactic cosmic rays;
    Plant physiologist and tree growth expert David Ellsworth, at the University of Western Sydney, in Australia, said it was an “intriguing phenomenon”, and that the hypothesis that the growth spurts were caused by diffuse radiation driven by galactic cosmic rays, was “reasonable”.
    I too think this may be a reasonable hypothesis and worthy of consideration.
    The study, published in the journal New Phytologist looked at the factors that influence the growth of Sitka spruce trees (Picea sitchensis) felled in the Forest of Ae in Dumfriesshire, Scotland.
    This particular tree species produces high amounts of BVOC’s, isoprene and monoterpines which can act as precursors for the formation of c.c.n.. Anything which promotes plant growth promotes the production of some addition material for cloud droplet formation, which must contribute to cloud formation and climate change.
    So we have Co2 enrichment , GCR , and then this;-
    If ocean plants are fertilised don’t they produce more DMS?
    This all seems to add to the factors which need to be accounted for when trying to establish the origins and actions of aerosols.

  10. 60
    Paul S says:

    Jeff Pierce – The range of station responses on that graphic is interesting. I thought it might be regional but, presuming ‘Newark’ refers to New Jersey, there doesn’t seem to be a well-defined pattern. Incidentally is there any reason why the Tibet record is much ‘cleaner’ than the others? Related to altitude?

    A completely different question: There seem to be a range of factors, relating to the composition of the atmosphere, which determine whether or not, and to what extent, GCRs can ultimately affect cloud formation. Is it possible that human emissions are changing atmospheric conditions in such a way that GCRs are more (or less) likely to have an effect?

  11. 61

    Hi Jos (47),

    Your question of why specifically low clouds would be affected by cosmic rays was addressed by Yu et al ( ) and I described it in an earlier RC post on cosmic rays and aerosol nucleation ( ):

    “Based on his model for ion induced nucleation, Yu found that at low altitude, the number of particles produced is most sensitive to changes in cosmic ray intensity. At first sight, this may be a surprising result in light of the increasing cosmic ray intensity with increasing altitude. The reason is that high aloft, the limiting factor for particle formation is the availability of sulfuric acid rather than ions. Above a certain GCR intensity, increasing ionization further could even lead to a decrease in ion induced nucleation, because the lifetime of ion clusters is reduced (due to increased recombination of positive and negative ions). In contrast, at low altitude particle formation may be limited by the ionization rate (under certain circumstances), and an increase in ionization leads to an increase in nucleation. “

  12. 62
    Hank Roberts says:

    For Jeff Pierce (or Gavin?) — Jeff above mentioned a possible Forbush event happening right now — can you capture the picture at (quickly before it updates) — for reference? Right now it shows the last few hours of Sept. 18th. (Or point to an archive if there is one). Seems like an area discussion will want to refer to again.

  13. 63
    Pete Dunkelberg says:

    Dr Pierce @ 58 Your comment ends with “This does mean that there will necessarily be large changes in clouds from changes in cosmic rays.”

    Is that as intended?

  14. 64
    CM says:

    Jim CA, re: bounding the possible effect of cosmic rays,

    Just as a coda to what Jeff Pierce said above, Erlykin et al. (2011) calculated a possible cosmic-ray contribution of 0.002 °C to the warming of the past 50 years.

    I’m sure they did this in all seriousness. I mean, I’m sure both the authors and the peer reviewers of the Journal of Atmospheric and Solar-Terrestrial Physics frown on the frivolous calculation of insignificant effects for sarcastic effect. But I still think it’s a funny way to say “no trend, no role”.

  15. 65
    Jos Hagelaars says:

    Thanks Bart for your answer in #61. I wasn’t aware of your posts here on RC, but will read them this evening. See you on

  16. 66
    Jeff Pierce says:

    #63 and #58: I meant to say, “this does NOT mean”, thanks Pete!

  17. 67
    Hank Roberts says:

    Looking at the link for Erlykin et al. the abstract sorts it out a bit:

    “… for the troposphere there is only a very small overall value for the fraction of cloud attributable to cosmic rays (CR); … probably ∼1% for clouds below ~6.5km but less overall. The apparently higher value for low cloud is an artifact.

    The contribution of CR to ‘climate change’ is quite negligible.”

  18. 68
  19. 69
    muoncounter says:

    J Gary Fox #42

    It is not uncommon in high energy particle physics for projects to take years to get funded, researched, designed and built. They could not have started taking real data until the LHC’s main beam was up and running stably (Sept 2008). Just look at the number of co-authors on Kirkby’s Nature paper; getting that many physicists to sit still long enough to agree on just the paper’s title could take months.

    No need looking for ‘political’ reasons.

  20. 70
    Paul S says:

    muoncounter – I don’t think the CLOUD experiment used the LHC. The supplementary information says they used something called the Proton Synchrotron.

  21. 71
    muoncounter says:

    You’re right. The Proton Synchrotron is part of the proton injector series for the LHC, effectively a medium energy ‘pre-accelerator’ stage. Substantial refitting was done on this older machine prior to LHC comissioning.

    CLOUD was approved and ran prototype testing in 2006 (see this 2007 status report and Duplissy et al 2010). That meant the earliest beam time they could get was 2007-2008. It was actually 2009 before CLOUD started taking data runs with the PS secondary beam.

    There is, as usual, no evidence of any ‘political attempt’ to slow the project. Rank speculation may get you somewhere on some blogs, but not here.

  22. 72
    Brian Dodge says:

    I downloaded GCR(neutron) data from [1], TSI anomaly data from [2], and HadCRU T anomaly(using the Defreitas et al “trick” of detrending) data from [3]. All data sets are monthly, and start in 1978.83. I normalized the pressure corrected GCR data by dividing by the average, and subtracting 1 from the results to create an anomaly series. Plotting TSI versus GCR yields a scatterplot with the expected negative correlation; GCR = -0.11*TSI, R^2=0.56. Plotting TSI versus HadCRU T yields a weak positive correlation; HadCRU T = 0.07*TSI, R^2=0.07. Since there is a negative correlation between TSI and GCR, there is also a negative correlation between GCR and HadCRU T; HadCRU T = -0.31*GCR, R^2=0.03. How do we sort out whether the temperature changes are driven by TSI, GCR, or a combination of the two?

    In the scatterplot of TSI versus GCR, there is not a perfect correlation – there are months when TSI anomaly and GCR anomaly are high, and months when TSI and GCR are both low – points in the first and third quadrants of the graph. There are 70 months when this is the case, from a span of 393 months. The TSI anomaly for these months ranges from -0.57 W/m^2 to 0.56 W/m^2, the normalized GCR anomaly ranges from -0.12 to 0.07 (-12% to +7%). The detrended and offset HadCRUT anomaly ranges from -0.37 to 0.36 degrees C.

    I isolated these months TSI, GCR, and HadCRU T, and looked at the correlations. Because I’ve restricted the data to those months when the TSI and GCR are positively correlated, that correlation is meaningless; however, if one looks only at TSI and GCR are both positive, or when both are negative, there are negative correlations; GCR = -0.02*TSI+0.03, R^2=0.02 when both are positive, GCR = -0.09*TSI – 0.06, R^2=0.2 when both are negative.

    Now, let us suppose that GCRs have a strong effect, causing most of the variation in global temperature by modulating clouds with the solar cycle. If the effect of GCR is dominant, the months when the GCRS are high, even when the TSI is high, temperatures should be lower; and when the GCRs are low, Temperature should be higher, even with lower TSI. Plotting HadCRU T versus TSI, only months when the TSI and GCR anomalies have the same sign yields a positive slope; HadCRUT = 0.11*TSI + 0.01, R^2=0.04. Plotting HadCRUT versus GCR gives a positive slope as well; HadCRU T = 0.62* GCR +0.01 for the full 70 months. The slope coefficient for the 20 months with the lowest GCR anomaly is -1.77, but the coefficient for the 20 months with the highest GCR anomaly is 3.76. Since the slopes for both TSI and GCR versus HadCRUT are positive, I conclude that although decreasing GCRs may enhance warming from concurrent increases in TSI, and vice versa, which is normally the case since they move in opposite directions, the influence of changes in TSI are the dominant factor. The negative correlation of GCR and temperature when the GCRs are negative, but overall positive correlation when both signs are analyzed may indicate a nonlinear effect, with stronger influences for decreasing GCR (such as the limited number of Forbush Decreases available for analysis).

    Wouldn’t all clouds over snow covered ground be warming, since the albedo is already high? And wouldn’t the increase in amplitude of variations of GCR with latitude (increasing towards the poles, where snow cover is common) tend to negate warming from low clouds near the equator because of this effect? Also, since the infrared radiation from the ground has a Lambertian distribution with its maximum intensity perpendicular to the surface, but the solar illumination is near parallel rays from the sun, a given cloud area at high latitudes will capture the same amount of outbound infrared, but reflect a smaller portion of the incoming solar radiation; at some angle where the sun is low on the horizon, this should cause low clouds to be warming.


  23. 73
    CM says:

    Hank #67, yes, bounding the fraction of cloud cover attributable to cosmic rays was the thrust of the Erlykin et al. paper, and I should have mentioned it, instead of trying to be funny.

    As for their calculation of the possible contribution of cosmic rays to global warming, it is unsurprisingly negligible for the posited 0.6% change in the mean intensity of cosmic rays over the past 50 years. It would be interesting to do the same estimate for the first half of the 20th century, though, given the 5% change Jeff Pierce mentioned above.

  24. 74
    J says:

    J Gary Fox is suspicious about the relatively long time it took before the cloud study was conducted. “Why the delay in conducting the needed experiment? The costs of the tests were not that high, about $10 million US.”

    That’s actually a pretty large price tag, and there’s a lot of competition for funds at that level. I was recently on a review panel evaluating approximately 60 proposals in the $3 million-$6 million range. The proposals had already been screened once, so these 60 represented just the good ones. Going into the panel, we were told that the agency would probably be able to fund 2 or 3 of the 60. As it turned out, there were so many good proposals in that group that the agency did some soul-searching and managed to come up with funds for 4-5 of them.

    In general, the question “how come proposal X hasn’t been funded?” is not a very useful question. Lots and lots of good proposals don’t get funded. It doesn’t mean there’s some conspiracy at work.

  25. 75
    dhogaza says:

    It doesn’t mean there’s some conspiracy at work.

    Not to mention that the notion that Gavin, working as a modeler at NASA GISS, could have a substantial impact on the research program and funding priorities at a high-energy experimental facility like CERN, with its many european collaborators involving a large percentage of the experimental physics community and a bunch of high-profile research groups, is rather ludicrous on the surface …

  26. 76
    muoncounter says:

    Jeff Pierce#55: Sorry for the slow response. Yes, the sun’s been spitting flares in the past week — see for details, as well as stunning aurora photos. The current FD seems to have bottomed at -5% (from Oulu’s neutron monitor current records); from the Dragic et al results, that’s not enough to make a measurable change in DTR.

  27. 77
    Jeff Pierce says:

    #72, Brian Dodge: Thanks for this analysis. Yes, it appears difficult to find connections between GCRs and temperatures in recent decades.

    I do think you are right that all clouds over snow will generally warm (there may be some exceptions to this depending on how “dirty” or how thick and aged it is). Interestingly, although the poles have the biggest fluctuations in cosmic rays, Marsh and Svensmark (2003) reported the biggest cloud changes with the 11-year solar cycle to be in midlatitude marine straticumulus clouds (although they did have a controversial correction to the ISCCP cloud data used in this paper in order to maintain the 11-year solar cycle).

  28. 78
    cbrock says:

    Nice article Jeff.

    What I’d like to see is a sensitivity calculation of particle nucleation and growth to all the variables that could affect it. It’s a very nonlinear process (or processes), strongly dependent upon sulfuric acid concentration, temperature, and the trace “contaminants”, as shown in the CLOUD study. But one of the primary controlling factors is preexisting particle surface area, which scavenges condensable gases before they have a chance to form new particles. And this surface area is largely governed by cloud formation and precipitation, which effectively removes particles to the surface, leaving very clean air behind. In my observations flying around the free troposphere, it is these regions, where preexisting particles have been scavenged, that are most likely to have newly formed particles present (other than concentrated industrial or urban plumes). So I have a hard time seeing how GCRs, which produce the ions that assist nucleation, play a controlling role when all these other processes are in play. Where rain/snow effectively scavenges the preexisting particles, nucleation tends to proceed; where it doesn’t, nucleation is more rare. At least in the free troposphere; there’s certainly a lot of evidence for nucleation occurring regularly over forested regions and in urban areas. But in these cases (e.g. the Finnish work) ion-mediated nucleation usually doesn’t dominate.

    Your thoughts?

  29. 79
    Jeff Pierce says:

    #78, cbrock: Thanks, Chuck! You definitely have a very informed perspective on aerosol nucleation and growth, so I’m glad you weighed in here. You are right that the global models that researches like me use don’t capture the nucleation near cloud outflow. I’ve always been curious about how important the near-cloud nucleation is for predicting aerosol indirect effects. It seems that Jan Kazil is making a serious attempt to look at this now with Graham Feingold. It would be nice if the 4 of us (and anyone else interested) could all chat when I’m in Boulder in November, especially since Jan is also an expert on ion-induced nucleation.

    My quick thoughts on your comment: I think you’re right that the 5-20% change in ion formation rate will play a secondary role to cloud scavenging in controlling nucleation near clouds (and then you still have issues with dampening in the nucleation-CCN-cloud connection). Furthermore, I think there might be a negative feedback loop for the effect of cosmic rays on near-cloud nucleation. An increase in cosmic rays would slightly increase nucleation, which would slightly increase CCN, which could slightly reduce precip, which would reduce aerosol removal, which would decrease nucleation. This is oversimplifying things, but its something to think about. I’m sure Jan and Graham would have more informed thoughts on this :)

  30. 80
    Jeff Pierce says:

    #60, Paul S: Sorry I forgot to respond before. Regarding your first set of questions about the various measurement sites, I don’t know for sure. I am not an expert on the cosmic ray measurements. muoncounter, can you weigh in here?

    Regarding your second question, yes, human emissions can definitely change the ability of cosmic rays to affect clouds. There are two basic competing effects (and some other complicating factors that I won’t touch on here): (1) Human-generated SO2 emissions have increased the amount of H2SO4 that forms in the atmosphere. The H2SO4 can increase the sensitivity of nucleation rates to cosmic rays and can increase the rate at which nucleated particles grow to become CCN. (2) Humans have increased the total amount of aerosol in general, so CCN concentrations are much higher. This reduces the sensitivity of CCN to changes in nucleation. Thus, the effect of cosmic rays on clouds through the clear sky mechanism may be different during pre-industrial times compared to today. However, my calculations show that there is a low sensitivity in both time periods.

  31. 81
    Septic Matthew says:

    this has been a good thread. my thanks to all.

  32. 82
    Jeff Pierce says:

    Yes, I agree with Matthew. Thanks everyone for great comments, questions and thoughts!

  33. 83
    Jeff Pierce says:

    Also, thank you Gavin for inviting me to write the post!

  34. 84
    lucien locke says:

    To Jeff Pierce and comments:

    Excellent presentation of cosmic ray/cloud connections. Informative, challenging, and in the case of #69-#71….entertaining, said with tongue firmly in cheek.

  35. 85
    HenkL says:

    Re the Dragic et al. paper.
    Most important issue is that there is no 1:1 relation between clouds and DTR. Sure, clouds have an important contribution to DTR, but there are other factors (only mentioned by Dragic). I used the data for a nearby station (Eelde, the Netherlands): only about 40% of the variation in DTR is ‘explained’ by variation in clouds.
    To calculate significance ‘t-statistics’ were used. Using data for Eelde, I found that for this station there is a random 29% chance to get 35 dates where the mean of DTR[d+4] – DTR[d] is greater then 0.38°C.
    Minor: the stations used in the paper are from Greenland to East-Siberian in the far east; a little bit more then European.

  36. 86
    David Lewis says:

    Richard Alley, during his 2009 AGU Bjerknes Lecture, pointed out that paleoclimate records provide a test of the hypothesis that cosmic rays are a significant influence on climate.

    He said 40,000 years ago the Earth’s magnetic field weakened in what is called the Laschamps anomaly, to about 10% of its current level, which allowed large amounts of cosmic radiation to enter the planetary system.

    He displayed this chart

    A video of his lecture is here:
    The section of his lecture where he discusses this issue starts around minute 42:05 in the video.

    Quoting from his lecture: “its a really interesting hypothesis. There’s really good science to be done on this. But we have reason to believe its a fine tuning knob…. …We had a big cosmic ray signal [points to the chart] and the climate ignores it. And its just about that simple. These cosmic rays didn’t do enough that you can see it.”