It's The Evidence, Stupid!

Guest Post by Willis Eschenbach

I hear a lot of folks give the following explanation for the vagaries of the climate, viz:

It’s the sun, stupid.

And in fact, when I first started looking at the climate I thought the very same thing. How could it not be the sun, I reasoned, since obviously that’s what heats the planet.

Unfortunately, the dang facts got in the way again …

Chief among the dang facts is that despite looking in a whole lot of places, I never could find any trace of the 11-year sunspot cycle in any climate records. And believe me, I’ve looked.

You see, I reasoned that no matter whether the mechanism making the sun-climate connection were direct variations in the brightness of the sun, or variations in magnetic fields, or variations in UV, or variations in cosmic rays, or variations in the solar wind, they all run in synchronicity with the sunspots. So no matter the mechanism, it would have a visible ~11-year heartbeat.

I’ve looked for that 11-year rhythm every place I could think of—surface temperature records, sea level records, lake level records, wheat price records, tropospheric temperature records, river flow records. Eventually, I wrote up some of these findings, and I invited readers to point out some record, any record, in which the ~ 11-year sunspot cycle could be seen.

Nothing.

However, I’m a patient man, and to this day, I continue to look for the 11-year cycle. You can’t prove a negative … but you can amass evidence. My latest foray is into the world of atmospheric pressure. I figured that the atmospheric pressure might be more sensitive to variations in something like say the solar wind than the temperature would be.

Let me start, however, by taking a look at the elusive creature at the heart of this quest, the ~11-year sunspot cycle. Here is the periodogram of that cycle, so that we know what kind of signature we’re looking for:

Figure 1. Periodogram, showing the strengths of the various-length cycles in the SIDC sunspot data. In order to be able to compare disparate datasets, the values of the cycles are expressed as a percentage of the total range of the underlying data.

As you’d expect, the main peak is at around 11 years. However, the sunspot cycles are not regular, so we also have smaller peaks at nearby cycle lengths. Figure 2 shows an expanded view of the central part of Figure 1, showing only the range from seven to twenty-five years:

Figure 2. The same periodogram as in Figure 1, but showing only the 7 – 25 year range. 

Now, there is a temptation to see the central figure as some kind of regular amplitude-modulated signal, with side-lobes. However, that’s not what’s happening here. There is no regular signal. Instead of there being a regular cycle, the length of the sunspot cycle varies widely, from about nine to about 15 years, with most of them in the 10-12 year range. The periodogram is merely showing that variation in cycle length.

In any case, that’s what we’re looking for—some kind of strong signal, with its peak value in the range of about 10-12 years.

As I mentioned above, when I started looking at the climate, like many people I thought “It’s the sun, stupid”, but I had found no data to back that up. So what did I find in my latest search? Well, sweet Fannie Adams, as our cousins across the pond say … here are my results:

Figure 3. Periodograms of four long-term atmospheric pressure records from around the globe.

There are some interesting features of these records.

First, there is a very strong annual cycle. I expected annual cycles, but not ones that large. These cycles are 30% to 60% of the total range of the data. I assume they result in large part from the prevalence of low-pressure areas associated with storms in the local wintertime, combined with some effect from the variations in temperature. I also note that as expected, Tahiti, being nearest to the equator and with little in the way of either temperature variations or low-pressure storms, has the smallest one-year cycle.

Other than semi-annual and annual cycles, however, there is very little power in the other cycle lengths. Figure 4 shows the expanded version of the same data, from seven to twenty-five years. Note the change in scale.

Figure 4. Periodograms of four long-term atmospheric pressure records from around the globe.

First, note that unlike the size of the annual cycle, which is half the total swing in pressures, none of these cycles have more than about 4% of the total swing of the atmospheric pressure. These are tiny cycles.

Next, generally there is more power in the ~ 9-year and the ~ 13-14 year ranges than there is in the ~ 11-year cycles.

So … once again, I end up back where I started. I still haven’t found any climate datasets that show any traces of the 11-year sunspot cycles. They may be there in the pressure data, to be sure, it is impossible to prove a negative, I can’t say they’re not there … but if so, they are hiding way, way down in the weeds.

Which of course leads to the obvious question … why no sign of the 11-year solar cycles?

I hold that this shows that the temperature of the system is relatively insensitive to changes in forcing. This, of course, is rank heresy to the current scientific climate paradigm, which holds that ceteris paribus, changes in temperature are a linear function of changes in forcing. I disagree. I say that the temperature of the planet is set by a dynamic thermoregulatory system composed of emergent phenomena that only appear when the surface gets hotter than a certain temperature threshold. These emergent phenomena maintain the temperature of the globe within narrow bounds (e.g. ± 0.3°C over the 20th Century), despite changes in volcanoes, despite changes in aerosols, despite changes in GHGs, despite changes in forcing of all kinds. The regulatory system responds to temperature, not to forcing.

And I say that because of the existence of these thermoregulatory systems, the 11-year variations in the sun’s UV and magnetism and brightness, as well as the volcanic variations and other forcing variations … well, they make little difference.

As a result, once again, I open the Quest for the Holy 11-Year Grail to others. I invite those that believe that “It’s the sun, stupid” to show us the terrestrial climate record that has any sign of being correlated with the 11-year sunspot cycles. I’ve looked. Lots of folks have looked … where is that record? I encourage you to employ whatever methods you want to use to expose the connection—cross-correlation, wavelet analysis, spectrum analysis, fourier analysis, the world is your lobster. Report back your findings, I’d like to put this question to bed.

It’s a lovely Saturday in spring, what could be finer? Gotta get outside and study me some sunshine. I wish you all many such days.

w.

For Clarity: If you disagree with someone, please quote their exact words that you disagree with. It avoids all kinds of pernicious misunderstandings, because it lets us all know exactly where you think they went off the rails.

Why The 11-year Cycle?: Because it is the biggest cycle, and we know all of the other cycles (magnetism, TSI, solar wind) move in synchronicity with the sunspots. As a result, if you want to claim that the climate is responding to say a slow, smaller 100-year cycle in the sunspot data, then by the same token it must be responding more strongly to the larger 11-cycle in the sunspot data, and so the effect should be visible there.

The Subject Of This Post: Please do not mistake this quest for the elusive 11-year cycle in climate datasets as an opportunity for you to propound your favorite theory about approximately 43-year pseudo-cycles due to the opposition of Uranus. If you can’t show me a climate dataset containing an 11-year cycle, your hypothesis is totally off-topic for this post. I encourage you to write it up and send it to Anthony, he may publish it, or to Tallbloke, he might also. I encourage everyone to get their ideas out there. Here on this thread, though, I’m looking for the 11-year cycle sunspot cycle in any terrestrial climate records.

The Common Cycles in Figures 3 and 4: Obviously, the four records in Figs. 3 & 4 have a common one-year cycle. As an indication of the sensitivity of the method that I’m using, consider the two other peaks which are common to all four of the records. These are the six-month cycle, and the 9-year cycle. It is well known that the moon raises tides in the atmosphere just as it does in the ocean. The 9-year periodicity is not uncommon in tidal datasets, and the same is true about the 6-month periodicity. I would say that we’re looking at the signature of the atmospheric tides in those cycle lengths.

Variable-Length Cycles, AKA “Pseudocycles” or “Approximate Cycles”: Some commenters in the past have asserted that my method, which I’ve nicknamed “Slow Fourier Analysis” but which actually seems to be a variant of what might be called direct spectrum analysis, is incapable of detecting variable-length cycles. They talk about a cycle say around sixty years that changes period over time.

However, the sunspot cycle is also quite variable in length … and despite that my method not only picks up the most common cycle length, it shows the strength of the sunspot cycles at the other cycle lengths as well.

A Couple of my Previous Searches for the 11-Year Sunspot Cycle:

Looking at four long-term temperature records here.

A previous look at four more long-term temperature records.

Atmospheric Pressure and Sunspot Data:

Madras

Nagasaki 

Tahiti to 1950  and Tahiti 1951 on (note different units)

Darwin to 1950  and Darwin 1951 on  (note different units)

Sunspots These are from SIDC. Note that per advice from Leif Svalgaard, in the work I did above the pre-1947 values have been increased by 20% to adjust for the change in counting methods. It does not affect this analysis, you can use either one.

For ease of downloading, I’ve also made up a CSV file containing all of the above data, called Long Term Atmospheric Pressure.csv

And for R users, I’ve saved all 5 data files in R format as “Long Pressure Datasets.tab”

Code: Man, I hate this part … hang on … let me clean it up a bit … OK, I just whacked out piles of useless stuff and ran it in an empty workspace and it seemed to fly. You need two things, a file called madras pressure.R  and my Slow Fourier Transform Functions.R. Let me know what doesn’t work.