Guest Post by Willis Eschenbach
In investigations of the past history of cosmic rays, the deposition rates (flux rates) of the beryllium isotope 10Be are often used as a proxy for the amount of cosmic rays. This is because 10Be is produced, inter alia, by cosmic rays in the atmosphere. Being a congenitally inquisitive type of fellow, I thought I’d look to see just how good a proxy 10Be might be for solar activity. Now most folks would likely do a search of the literature first, to find out what is currently known about the subject.
I don’t like doing that. Oh, the literature search is important, don’t get me wrong … but I postpone it as long as I possibly can. You see, I don’t want to be mesmerized by what is claimed to be already known. I want to look whatever it is with a fresh eye, what the Buddhists call “Beginner’s Mind”, unencumbered by decades of claims and counter-claims. In short, what I do when faced with a new field is to go find some data and analyze it. After I’ve found out what I can from the dataset, and only then, do I search the literature to find out what other folks might believe. Yes, it costs me sometimes … but usually it allows me to find things that other folks have overlooked.
In this case, I found a gem of a dataset. Here is the author’s summary:
Annually-resolved polar ice core 10Be records spanning the Neutron Monitor era
Abstract: Annually-resolved 10Be concentrations, stable water isotope ratios and accumulation rate data from the DSS site on Law Dome, East Antarctica (spanning 1936-2009) and the Das2 site, south-east Greenland (1936-2002).
The only thing better than data is recent data, because it is more likely to be accurate, and here we have seven decades of recent 10Be deposition rates (fluxes). So, without fanfare, here’s the data in question
So … what’s not to like about these records? Well … lots of things.
The first unlikable item is that the correlation between these two 10Be datasets is pathetic, only 0.07. Seems to me like this would be enough in itself to put the whole 10Be—cosmic rays connection into doubt. I mean, if the two best recent dataset don’t agree with each other, then what are we supposed to believe?
The next problem is even larger. It is the lack of any clear 11-year signal from the variations in cosmic rays. It is well-known that cosmic rays are deflected from the solar system by the magnetic field of the sun, which varies in general sync with the sunspots. As a result, the numbers of cosmic rays, and presumably the 10Be flux rates, vary in an 11-year cycle inversely to the sunspot cycle. Here’s what the relationship looks like:
Figure 2. Sunspots and cosmic rays (as indicated by the neutron count). SOURCE
So the relation between cosmic rays and sunspots is quite solid, as you can see above. However, the problem with the 10Be records in this regard is … they have no power in the 11-year cycle range. Sunspot data has power in that range, as does the neutron count data representing cosmic rays … but the 10Be data shows nothing in that range. Here’s the periodicity analysis (see here et seq. for details of periodicity analyses):
As you can see, we have no power in either the 11-year or 22-year bands … and if you look at Figure 1, you can see that their correlation with the sunspots is … well … pathetic. The correlation between Greenland 10Be and sunspots is -0.10, and between Antarctica 10Be and sunspots is even worse, -0.03 … like I said, pathetic. A cross-correlation analysis shows slightly greater correlations with a 2 year lag, but not much. However, the lack of the 11-year peaks periodicity analysis (or visible 11-year peaks in the 10Be data) suggests that the lag is spurious.
The problem is, both the sunspots and the cosmic ray counts have a huge peak in periodicity at 10-11 years … but the 10Be records show nothing of the sort.
So, at this point I’m in as much mystery as when I started. We have two beryllium-10 records. They don’t agree with each other. And according to both periodicity and correlation analysis, they don’t show any sign of being connected to anything related to the sunspots, whether by way of cosmic rays, TSI, or anything else …
Now that I’ve finished the analysis, I find that the notes to the dataset say:
Cosmogenic 10Be in polar ice cores is a primary proxy for past solar activity. However, interpretation of the 10Be record is hindered by limited understanding of the physical processes governing its atmospheric transport and deposition to the ice sheets. This issue is addressed by evaluating two accurately dated, annually resolved ice core 10Be records against modern solar activity observations and instrumental and reanalysis climate data. The cores are sampled from the DSS site on Law Dome, East Antarctica (spanning 1936–2009) and the Das2 site, south-east Greenland (1936–2002), permitting inter-hemispheric comparisons.
Concentrations at both DSS and Das2 are significantly correlated to the 11-yr solar cycle modulation of cosmic ray intensity, r = 0.54 with 95% CI [0.31; 0.70], and r = 0.45 with 95% CI [0.22; 0.62], respectively. For both sites, if fluxes are used instead of concentrations then correlations with solar activity decrease.
If you use flux rates the “Correlations with solar activity decrease”??? Yeah, they do … they decrease to insignificance. And this is a big problem. It’s a good thing I didn’t read the notes first …
Now, my understanding is that using 10Be concentrations in ice cores doesn’t give valid results. This is because the 10Be is coming down from the sky … but so is the snow. As a result, the concentration is a factor of both the 10Be flux and the snow accumulation rate. So if we want to understand the production and subsequent deposition rate of 10Be, it is necessary to correct the 10Be concentrations by using the corresponding snow accumulation rate to give us the actual flux rate. So 10Be flux rates should show a better correlation with sunspots than concentrations, because they’re free of the confounding variable of snow accumulation rate.
As a result, I’ve used the flux rates and not the concentrations … and found nothing of interest. No correlation between the datasets, no 11-year periodicity, no relationship to the solar cycle.
What am I missing here? What am I doing wrong? How can they use the concentration of 10Be rather than the flux? Are we getting accurate results from the ice cores? If not, why not?
These questions and more … please note that I make no overarching claims about the utility of 10Be as a proxy for sunspots or cosmic rays. I’m just saying that this particular 10Be data would make a p-poor proxy for anything … and once again I’m raising what to me is an important question:
If the 10Be deposition rate is claimed to be a proxy for the long-term small changes in overall levels of cosmic rays … why is there no sign in these datasets of it responding to the much larger 11-year change in cosmic rays?
I have the same question about cosmic rays and temperature. There is no sign of an 11-year cycle in the temperature, meaning any influence of cosmic rays is tiny enough to be lost in the noise. So since temperature doesn’t respond to large 11-year fluctuations in cosmic rays, why would we expect temperature to track much smaller long-term changes in the cosmic ray levels?
Always more questions than answers, may it ever be so.
My regards to everyone, guest authors, commenters, and lurkers … and of course, Anthony and the tireless mods, without whom this whole circus wouldn’t work at all.
COMMENTS: Please quote the exact words that you are referring to in your comment. I’m tired of trying to guess what folks are talking about. Quote’m or you won’t get traction from me. Even if the reference is blatantly obvious to you, it may not be to others. So please, quote the exact words.
DATA: 10Be original data, Excel spreadsheet
CODE: Just for fun, I’ll put it here to show how tough this particular analysis was:
source("~/periodicity functions.R") par(mgp=c(2,1,0),cex.axis=1) spotsraw=ts(read.csv("monthly ssn.csv")[,2],start=c(1749,1),frequency=12) Annual.Sunspots=window(aggregate(spotsraw,frequency=1,FUN=mean),start=1937,end=2009) plot(Annual.Sunspots) theflux=ts(read.csv("Polar 10Be Flux.csv")[,2:3],start=1937,frequency=1) theoxy=ts(read.csv("Polar 10Be Flux.csv")[,4:5],start=1937,frequency=1) plot(cbind(theflux,theoxy)) fulldata=cbind(theflux[,1],theflux[,2],Annual.Sunspots) colnames(fulldata) = c("Greenland 10Be Flux","Antarctica 10Be Flux","Sunspots") plot(fulldata,main="",yax.flip=TRUE) title(main="10Be Flux Rates in Greenland and Antarctica\n(atoms / square metre / second)", line=1,cex.main=1.1) cor(ts.intersect(fulldata),use="pairwise.complete.obs") periodsd(theflux[,1],doplot=TRUE,timeinterval=1,add=FALSE,col="blue3", maintitle="Periodicity Analysis, Ice Core 10Be Flux\nGreenland (blue) and Antarctica (red)") periodsd(theflux[,2],doplot=TRUE,timeinterval=1,add=TRUE,col="green3")
You’ll need the code for the periodicity functions, it’s here.