Falsification tests of climate hypotheses
Cosmic rays and clouds at various latitudes
An exchange with Prof. Terry Sloan of Lancaster University
I’m promoting to the start of a new post a comment on an earlier post that came from Terry Sloan, together with my reply and his comment on my reply. I’ve included a graph that he sent in an e-mail because it wouldn’t upload into the Comments section.
After that, the discussion continues here with further remarks from me.
Sloan is one of the severest critics of the Svensmark hypothesis that cosmic rays influence the Earth’s low clouds. The earlier post, entitled “Do clouds disappear when cosmic rays get weaker?”, was concerned chiefly with whether or not sudden changes called Forbush decreases have observable effects on cloud cover. You can see that post in full here: http://calderup.wordpress.com/2010/05/03/do-clouds-disappear/
But the present interaction with Sloan mainly concerns a different question, about the influence of the Earth’s magnetic field. To help readers to get quickly up to speed, here’s the most relevant extract from my original post:
Bringing the Earth’s magnetism into the story
Both Wolfendale’s and Arnold’s teams repeated a different complaint going back to Sloan and Wolfendale in 2008, about clouds at different latitudes. In promising to return to it later, I was saving the neatest rebuttal till last. Here, for example, is how Laken, Wolfendale and Kniveton expressed their concern.
“An analysis of the latitudinal distribution of the [low cloud fraction] variations reveals that this decrease is predominately located at mid to low latitudes, whereas if the phenomenon were related to variations in the [cosmic ray] flux it should be predominately located at high latitudes.”
What this is all about is the influence of the Earth’s magnetic field on the influx of cosmic rays. As a shield of sorts, it works much better in the tropics than towards the magnetic poles. So with more cosmic rays coming in at higher latitudes, and varying more, shouldn’t there be a bigger effect on clouds there, than at at low latitudes?
The argument was most clearly illustrated by Sloan & Wolfendale, with a diagram that related not to sudden Forbush decreases but to long-term variations in cosmic rays over Solar Cycle 22 (1983-96) and to an earlier report on their effects on clouds by Nigel Marsh and Henrik Svensmark in 2000. I’ve replaced rather technical labels on the axes of the graph with simpler ones of my own.
The line NM is the variation in cosmic rays at different latitudes, which Sloan & Wolfendale say should be followed by any variation in low cloud cover if the Svensmark effect is real. The various symbols show the actual variations in the clouds according to an analysis of ISCCP data, which are obviously missing the target line completely.
Test failed? Not a bit of it. Here is the same graph with a red line added by Svensmark, showing how he computes that the cloud effect should vary with latitude. It fits the Sloan & Wolfendale data surprisingly well, when you remember how “noisy” the ISCCP data are.
Test passed. We have come to the nub of the misconception, where the critics haven’t grasped an elementary point about Svensmark’s physics. For ten years he has said the clouds most affected by cosmic rays are low clouds. So the cosmic rays that matter are charged particles (mainly muons, heavy electrons) that penetrate low into the atmosphere. They’re generated mostly by very energetic protons from the Galaxy on which the Earth’s magnetic field has little influence. Hence the much reduced slope of the red curve, compared with Sloan & Wolfendale’s NM slope.
NM stands for neutron monitors, and there’s the blunder. Neutrons are very handy for showing changes in cosmic ray intensities, whether in a Forbush decrease or during a solar cycle. But as high-school students know, neutrons are uncharged. They don’t ionize the air or affect the clouds. To rely on neutrons to tell you what the clouds should do is as rash as expecting tigers to have smoke coming out of their heads.
The overriding importance of the muons is the reason why Svensmark’s team went to so much trouble to compute the ionization of the lower air for many Forbush decreases. None of the critics cited here used the same reckoning to lead them to the effects on clouds. This is what lay behind Svensmark’s remark to the BBC back in 2008, that Sloan “simply failed to understand how cosmic rays work on clouds”
The extract from the earlier post ends, which brings us to the ongoing exchange of views.
Sloan’s initial comment (27 July 2010)
Sorry I have only just seen this piece.
Henrik Svensmark is wrong to assume that most of the ionization at cloud forming height is caused by cosmic ray muons. There is a significant contribution from the so called soft component of cosmic rays. The ionization from muons and the soft component then produces a latitude dependence of the ionization which follows roughly the shape of the neutron monitor (NM) data. In our paper only the shape of the distribution was used (not the absolute level) to show that the distribution was mainly flat with less than 23% attributable to a shape like the NM data. Hence less than 23% of the dip in the globally averaged low cloud cover, reported in the original Marsh and Svensmark paper, could be caused by cosmic ray ionization. This is now borne out by the following solar cycle (peaking in 2000) where there is little sign of a dip in the cloud cover. If changes in cosmic rays cause significant changes in cloud cover we should have seen a dip in 2000 similar to the one seen in 1990 (see http://isccp.giss.nasa.gov/climanal7.html ).
There is another argument which shows that it is inconsistent to postulate that most of the ionization comes from muons and at the same time that these cause the dip in cloud cover seen in 1990. It goes like this.
Let us suppose that Henrik is correct and the change in cloud cover seen in 1990 is caused by the solar modulation of ionization from muons. This has been measured to be 2-3% (Ahluwalia, JGR 102 (1997) 24229). The dip in cloud cover at the 1990 solar maximum shown in the Marsh and Svensmark paper represents a fractional change in cloud cover of 4.5%. For a 2-3% change in ionization from muons to produce a 4.5% change in cloud requires that the change in cloud cover varies roughly as the square of the change in ionization rate. Theories of this process would predict that such a variation would be more like a square root behaviour rather than a square. Hence it is inconsistent to postulate that changes in ionization from muons alone cause the change in cloud cover seen in 1990. None of this says that there is no effect of cosmic rays on clouds. However, such an effect can only be a minor one.
Calder’s reply (29 July 2010)
Thanks for your comment, Terry.
1) Concerning muons versus the soft component, most of the low clouds that matter in the Svensmark hypothesis are oceanic clouds ~1 km above sea level. There, most of the ionization is indeed caused by cosmic ray muons. The CRC Handbook of Chemistry & Physics tells me that at sea level the soft component contributes about one third of the ionization and at low altitudes scarcely more. In the same elementary reference book I see that at sea-level the flux (hard and soft) increases by about 7.5% from the Equator to ~latitude 40 and then flattens out. This is very like the observed variations of low clouds in your NM plot. So the neutron monitors featured in that plot are an extremely poor guide either to the magnitude or the the shape of latitudinal variation in low-altitude ionization. The main reason for the muted latitudinal effect is that nearly all low-altitude muons are created by primary cosmic rays so energetic that the geomagnetic field has no effect on them. See Fig. 6 on page 60 of The Chilling Stars. If you want to check this, the CORSIKA program of the Karlsruhe Institut für Kernphysik is available at http://www-ik.fzk.de/corsika/ — but be warned, doing this physics is time-consuming. By the way CORSIKA takes full account of the soft component as well as the muons.
2) Concerning cloud behaviour in “the following solar cycle” (i.e. Cycle 23 peaking ~2000) the ISCCP D2 IR low cloud data that you cite have well-known problems arising from the changing population of weather satellites and changes in viewing angle that affect the low cloud assessments. See for example Amato T. Evan, Andrew K. Heidinger and Daniel J. Vimont, “Arguments against a physical long-term trend in global ISCCP cloud amounts”, Geophysical Research Letters, Vol. 34, L04701, 2007. When those problems have been sorted out, I expect that the downward trend in low cloud cover will disappear and the de-trending will also reveal a variation in low cloud cover in Cycle 23 very similar to that in Cycle 22. We’ll see, when the physics is done.
3) On your last point, about proportionality, the Ahluwalia data show a 4% fall in the muon count at the time you mention, not 2-3% as you suggest. See Fig. 2 in Svensmark, Physical Review Letters, Vol. 80, pp 5027-30, 1998. The effect on clouds may indeed be proportional to the square root of the ionization density, but if I write Δ(LCC)=k.Δ√(muon change), a 4% muon change could give a 4.5% change in LCC if k=2.25. The existence of a constant k would be unsurprising in view of the complex atmospheric chemistry involved in translating ionization into cloud condensation nuclei. But in any case, uncertainties about the ISCCP data, mentioned in (2), and questions, for example, about the atmospheric distribution of sulphuric acid molecules, mean that it would be rash to try to establish a power law from a single pair of data.
Sloan’s second comment (29 July 2010)
The soft component is generated by pi-zero decay and is the same over both land and sea.
I have been studying the simulations and data and the latitude dependence of the solar modulation of cosmic ray ionization resembles that of neutron monitors. The modern simulations give a bigger soft component than is given in the CRC handbook. (I will send you a plot – I do not know how to put it on here).
(Inserting Sloan’s e-mailed plot and his accompanying explanation)
“The GEANT4 simulation by the Berne group which shows almost equal soft (electron) and muon component near sea level [=extreme right]. This squares with my experience in testing scintillators and particle detectors (I am an experimental particle physicist). There are always many events that do not have muons in them (but this is subjective).”
(Sloan’s second comment concludes)
Concerning solar cycle 23 – if you use the data through until 2009 you can see that there is little sign of a dip. Henrik’s detrending only used data just beyond 2000 and this made it look like a dip. Have a look at the link to the ISCCP web site – the plots are there. There is not much sign of s dip in solar cycle 23. However, we will see when they have recalibrated.
The Ahluwalia paper gives 3.6% as the fall in the muon rate in solar cycle 22. This looks anomalously large compared to other solar cycles so is probably a statistical fluctuation. However, let us take this value. Calculus shows us that if a change of 4.5% in cloud cover comes from a change of 3.6% in ionization one needs cloud cover to vary as the power 4.5/3.6=1.25 whereas simple square root behaviour would would require a value 0.5 here. Your constant k does not enter since we are discussing fractional changes i.e. relating Delta(LCC)/LCC and Delta(Ionization)/Ionization (i.e. assuming LCC=k Ion^power).
Continuing afresh – reply by Calder (9 August 2010)
[Preamble] Forgive the slow reaction, Terry. I’ve been digging deep into some history about latitudinal effects, and into muons versus electrons at low altitudes, including the surprising GEANT4 ATOMOCOSMICS plot that you’ve sent. (Sorry you had difficulty inserting it in your comment — I’ve done that here).
Starting with your concluding remarks, I withdraw my constant k, which was a blunder. Your challenge about an apparently peculiar power law for cloud cover remains on the table. So does the question of what the recent ISCCP data will or won’t show when the needed corrections for viewing angles have been made.
Otherwise, this reply continues with the theme of cosmic rays at high latitudes and low altitudes, and with contesting your claim (offered as a falsification of the Svensmark hypothesis) that any variations in low cloud cover over a solar cycle should be much larger at high geomagnetic latitudes if the hypothesis were correct.
What has always puzzled me about the “NM” graph in Sloan & Wolfendale 2008 is the belief that latitudinal changes in the ionizing component of the cosmic rays should conform with the rapid increase in neutron counts. Neutrons are produced copiously at high altitudes and high latitudes, and are much more durable in penetrating the atmosphere than charged particles are. The cosmic rays that ionize the air at low altitudes tell a quite different story.
As an undergraduate half a century ago (when cosmic ray research was second in importance only to nuclear physics) I was assured by the Cambridge experts of the time that the ionizing component versus latitude followed an S-shaped (sigmoid) curve. That’s what I’ve been excavating from the literature.
A classic study at sea level was done by Milliikan and Neher (1936). Robert Millikan, remember, was the physicist who coined the very name “cosmic rays” in 1923, the year when he also won the Nobel prize for measuring the charge on the electron. Neher was an instrumental wizard. The reference is R.A. Millikan & H.V. Neher, Physical Review, Vol. 50, pp. 15ff., 1936, and the full text is available at http://authors.library.caltech.edu/6680/
Millikan and Neher installed detectors of cosmic-ray ionization in ocean liners, to record the changes as the ships travelled through the geomagnetic latitudes. The detectors were usually shielded with lead 10 centimetres thick. But because of the importance Sloan suggests for “soft” electrons (in his second comment) I’ve selected a figure from the paper using data from unshielded detectors. Any special influence of soft electrons would have had every chance of to show up, but it didn’t.
The now-conventional sigmoid curve is plainest to see on the right. The authors comment that the results from the unshielded instrument were more erratic than from their usual shielded instrument because less care was taken in keeping passengers well away. (They were criticizing the ships’ officers who tended the experiments, and were more careful with shielded detectors.)
Later, Neher and colleagues flew a cosmic-ray telescope in a borrowed B-29 bomber, following chosen routes to see how the latitude effect looked at 30,000 feet (an altitude of about 9 km). This experiment also had the benefit of speed, so that the data were not compromised by possible solar-induced changes in cosmic radiation from week to week, as was possible on long sea voyages. The resulting paper by Biehl, Neher and Roesch, Physical Review, Vol. 76, pp. 914 ff. 1949 is available at http://authors.library.caltech.edu/8698/1/BIEpr49b.pdf
Again I choose results where there was no shielding on the telescope to cut out soft electrons. For the uppermost curve the telescope pointed vertically upwards; in the others it was tilted 45 degrees to west or (lowest) 45 degrees to the east.
Three nice sigmoids, nothing like the Sloan-Wolfendale NM straight line, and this at an altitude much higher than Svensmark’s low clouds. Throughout the troposphere, where the weather is made, there is virtually no increase in ionizing cosmic rays beyond 45 deg. magnetic latitude.
But the percentage increase seen in the sloping part of the curve is greater at this higher altitude. The effect of altitude is apparent in a fuller version of the Svensmark 2008 graph that I showed earlier in rebuttal of the Sloan-Wolfendale NM graph.
This is an adaptation of the original Sloan-Wolfendale figure, without the simplified labelling of axes used in my earlier versions. High geomagnetic latitudes are on the left, and low on the right. The NM (neutron monitors) line is what Sloan & Wolfendale say the low cloud variations in a solar cycle should obey if the Svensmark hypothesis is right, and the symbols show low cloud data failing to conform. The lower curves are Svensmark’s own computations using the Karlsruhe CORSIKA program, showing the variations in cloud cover expected through the solar cycle at various low altitudes. Notice how, in each case, the effect flattens at high latitudes (on the left) in a manner reminiscent of the observational sigmoid curves of ionization shown earlier.
Sources of the low-altitude ionization
To persevere with this cosmic-ray physics, which still seems to be eluding Sloan, let’s see where the low-altitude cosmic rays come from. Here is a plot of the major ionizing components of the cosmic rays at different altitudes from a 1984 textbook, available via the NASA Astrophysics Data System.
And to pursue the source for the low-altitude cosmic rays, here’s a recent computation with the CORSIKA program. It’s for the ionization at different altitudes resulting from the impact of energetic (100 GeV) protons on the atmosphere. Source: P.I.Y. Velinov et al., Advances in Space Research , Vol. 44, pp.1002–1007, 2009
Here muons are exceptionally influential in ionization at low altitudes (on the right). “EM ionization” is the source of soft electrons. A similar plot by Velinov for 10 GeV protons allows a somewhat larger role for soft electrons at sea-level, but with a total ionization two orders of magnitude less.
As for 1 GeV protons, which are representative of the extra cosmic rays admitted by the geomagnetic field at high geomagnetic latitudes, they have no influence at all at the lowest altitudes. Here’s the Velinov plot for 1 GeV.
The ionization simply fades away by about 750 g cm-2, or 2650 metres altitude. This shows why the increased cosmic-ray influx at high latitudes has little effect at low altitudes – the nub of the problem with the Sloan-Wolfendale NM plot.
A remaining technical question is why the GEANT4 plot in Sloan’s second comment (graph repeated below) seems to contradict everything else said here, and to demote the muons and boost the electrons at low altitudes (again on the right).
The clue to the figure’s peculiarity is the curve labelled “Balloon experiment”. Going back to the report that is the source of the plot (L. Desorgher et al., International Journal of Modern Physics A, Vol. 20, pp. 6802-04) here’s the full caption to Figure 1, of which Sloan’s offering is the left panel. :
Fig. 1. The solid lines represent the ATMOCOSMICS computed flux of cosmic ray shower particles vs atmospheric depth over Moscow during the maximum of solar activity. The dotted line in the left panel represents the year 2000 averaged flux of cosmic ray measured over Moscow by the balloon experiment from the Lebedev Physical Institute. This experiment is sensitive to fluxes of electrons with energy > 200 keV, protons with energy > 5 MeV, muons, and 1% of gamma rays with energy > 20 keV. The highest solid line in this panel represents the total flux of these particles computed with ATMOCOSMICS.
So this plot that Sloan wants to introduce into a falsification test of the Svensmark hypothesis, while asserting that “The modern simulations give a bigger soft component than is given in the CRC handbook”, is no general assessment of cosmic rays. It’s a special calculation related to the instrumentation of a particular balloon experiment over Moscow ten years ago, which happened to be more sensitive to soft electrons than to other low-energy cosmic ray particles.
For that quite entertaining reason, it is wholly irrelevant to an attempt to challenge the Svensmark hypothesis by invoking the role of soft electrons. Falsification attempt still fails and the old handbooks and textbooks are still trustworthy.