Home About My Books Résumé Chronological

Sunspot Minima

Created/Modified: 2020-11-22/2020-11-23

Are we headed into the kind of major solar minimum that gave us the horrors of the Little Ice Age in the 1600s?

There is (some) reason to think so.

Fortunately, maybe, the net effect this time around will be to buy us a bit more time to get our carbon habit under control... although if recent history is any guide we will have a good chance of wasting that opportunity.

Before asking "What are the implications for <<currently trendy group of interest>>?" it's worth asking, "Is this even likely to happen?"

Most punditry can't even get the past right. Here's a typical pundit's claim, which is a mixture of irrelevancy and falsehood: "The most confident pre-election pollsters proclaimed themselves 99 percent certain of the result that didn’t happen. Even the least confident predicted exactly what did not occur.".

What the confidence of a person has to do with the expected accuracy of their prediction is obscure, since there is good evidence that a large class of cases exists where confidence and accuracy are anti-correlated.

Nor did "the least confident prediction" of the 2016 election outcome (presumably 538) "predict exactly what did not occur". They gave a 30% chance of what occurred occurring, which... checks notes... happens about 30% of the time. It isn't even clear what "least confident" means in this context, because Nate Silver was highly confident in the 538 model predictions, but the model had high uncertainty given the data.

Replacing data with bias--which is what ideology does--is not a recipe for improved predictive accuracy, however much you admire Thucydides as an astute observer and commentator on human behaviour (which in my case is: quite a lot, really).

Given we often can't even get right what was predicted about the past, it's important to look carefully at what is being predicted today about the future.

Knowledge is inherently uncertain, and accurate predictions accurately reflect this uncertainty. When examining predictions we should be less concerned about "right" and "wrong" than "properly managed uncertainty": do they, over time, give us probabilities that reflect actual happenings? If we look at all the candidates 538 gives a 30% chance of winning, do they in fact win about 30% of the time? They do, so we have strong evidence that 538 is managing uncertainty well, and we should trust the methods they use to continue to do so in future.

So my goal here is to evaluate the uncertainty in the prediction, "We are entering a grand solar minimum in the next decade, which will plausibly have a significant effect on the Earth's climate."

I'm not going to say if it's "true" or not, because that's not even meaningful. It's meaningful to ask ten years from now "How close did it turn out to be to measured reality?" but not "Was it true?" There is no magic measure of closeness that tips it over from one binary category ("false") to another ("true"). There's a continuum of closeness to measured reality. If it's miles off we can informally say it's "false", if it's really close we can informally say it's "true", but if we reify those terms of convenience--"true" and "false"--into Objectively Real Categories we're making a big mistake that will come back to haunt us.

There are people who get upset about this and insist I'm answering the wrong question, that they want to know what will happen, not the various odds on different things that might happen. To those people I've gotta say: sorry, can't help. Neither can anyone else, and you shouldn't trust people who say otherwise. They're selling snake oil.

So... what are the odds? Is Professor Zharkova some kind of kook? Her stuff is being promoted by the "electric universe" people, who are all quite mad, but guilt by association is not a good look, really. It's also been glommed onto by some climate change denialists, and she has contributed a bit to that as well. That doesn't impact the correctness of the solar physics, though.

I've read the paper that this editorial is based on, and while I think the top-line claims are over-stated, the result is not completely crazy regardless of what external considerations come into play. But there is a big gap between "not completely crazy" and "anything close to certainty." I'd say this falls in the lower range of that gap.

The argument has multiple parts starting with these initial ideas:

  1. The sun is a complex object whose heat comes from nuclear fusion in the core, but the heat is transported to the surface--and radiated away as light--in a variety of ways, some of which are mediated by the solar magnetic field.
  2. Sunspots are visible manifestations of the interaction of the solar magnetic field with the hot plasma (conducting gas) that is the sun's "atmosphere", the outer layer that the internal heat has to pass through to get to, well, everything, including the Earth.
  3. Sunspots undergo a well-known eleven year cycle that has been measured over the past four hundred years. At each minimum there can be months or even years with no sunspots whatsoever.
  4. Based on historical climate data, the number of sunspots at the minimum is correlated with solar irradiance. That is, how much energy the sun puts out varies by a small amount as the number of sunspots increases and decreases.
  5. The normal 11 year cycle is too fast to have a significant impact on climate, but there is evidence of a longer 300-400 year cycle that does. The last "grand solar minimum" of this longer cycle is called the Maunder Minimum, and overlaps the global cooling episode known at the Little Ice Age, in the 1600s. Which may be a coincidence.

All of that is pretty straightforward stuff. The new analysis takes off from it as follows:

  1. The solar magnetic field is extremely complex and not very well understood, but we now have some very good whole-sun measurements of it covering four full sunspot cycles.
  2. These data can be processed using Principle Components Analysis (PCA) to find the major modes in solar magnetic field variation.
  3. This analysis reveals two major modes with slightly different frequencies that account for almost 40% of the variation in solar magnetic field over the period of observation.
  4. These modes can be modelled with a relatively simple mathematical form, and the analysis of these forms show that the two modes can interfere with each other to produce both the observed 11 year variation as well as a long-term "beat frequency" of about 350 years in the overall strength of the solar magnetic field.
  5. This mathematical model, when projected back into the past, predicts the Maunder Minimum, 400 years ago.
  6. This mathematical model, when projected forward into the future, predicts a "Modern Minimum" of comparable scope to the Maunder Minimum that we are just about to enter.

So... how plausible is all that?

First off: I am not a solar physicist or a plasma physicist. I had to look up what "poloidal field" meant when reading this stuff (it's a field that results from a closed loop that starts off lying in a plane being twisted out of plane, so after the first half-turn it looks like a figure-eight, and after a bunch of twisting and rotations ends up looking like something from a Spirograph.)

So my comments are not aimed at the details of the physics, which I'm not competent to address. Although the physics are so uncertain--which I'll discuss below--that this is not as big a deficit as one might think.

What I'm doing here is walking through how a scientist looks at things that are not close enough to their own field to evaluate directly. But there are 'tells' you look for, and these are some of mine, based in part on the areas of the work--like PCA, and model analysis generally--that I am an expert in.

PCA is a relatively conventional kind of analysis that in most cases pulls out robust and (sometimes) "natural" features from complex datasets, which makes them easier to think about. These features tend to represent the underlying physics in interesting ways, but should still be treated with caution, because they are ultimately just artifacts of a purely mathematical procedure. They could be physically meaningless.

In this case, there is an argument--handwavey, but still an argument--that the two major principle components are the result of two different layers in the solar envelope interacting with each other. The solar magnetic field is due to some kind of dynamo effect in which "meridonal currents" that flow from the solar equator toward the poles an generate all the magnetic phenomena we see. While the general picture is clear, the details are not, Zharkova and her collaborators posit two layers of meridonal currents that interact to produce the two major principle components (one of which seems to be about ten times more powerful than the other...)

While this doesn't look completely crazy to me, the thing about one component being ten times smaller than the other bothers me. It appears, in fact, that the authors have divided the components by their eigenvalues (the eigenvalue is a measure of the "size" of the component in an important sense) to give them the same scale. This is not justified by anything.

They do this analysis over four solar cycles, and get similar results if the restrict it to any two or three cycles, which is good: it suggests that their results are mathematically robust.

They then fit the components--which only explain 40% of the solar variance over the four cycles they have data for--and extrapolate those fits into the past. This reproduces something that looks like the Maunder Minimum at around the right time. If they then project this model into the future, they get an upcoming grand solar minimum.

There are several issues I have with this. The big one is: where does their model come from? They have two components with time-varying frequencies and different functional forms, and near as I can tell no particular grounding in the underlying physics. Nor are the functional forms all that simple, mathematically. I could totally get behind doing a Fourier analysis of the components they extract, on the basis that some combination of periodic functions has to go into them, and if that's all you know, then Fourier is what you do. They instead choose something quite different: a product of cosines with one of them having sinusoidal variation in frequency.

They have a paper describing their procedure for fitting the principle components but I confess to not understanding the underlying logic that produced the functional form they chose to use for their fit components.

What I do know is a) it's physics-free and b) it has a lot of parameters... 60, by my count.

A "model" that is a fit to the data with a high number of parameters is not a good tool for extrapolation. I've been critical of GCMs (general circulation models, which are what are used to study global climate) on precisely this basis and they have a much better grounding in physics and far fewer free parameters of any importance. They have, unsurprisingly, been badly wrong in detail--under-predicting warming at high latitudes, in particular--while getting the general picture correct (it's hard not to, as that is a matter of basic energy balance).

Never-the-less, it is intriguing that the model does predict the Maunder Minimum or something like it. One does wonder, though: how many other variations on this kind of PCA extraction and fitting procedure did they try before they found one that fit the expected past?

If that number was more than "one or two" then the significance of fitting a single major feature in the past is vastly diminished. Unfortunately, they do not tell us about all the stuff that didn't work, only the stuff that did. If there is very much of the former then what we are seeing is nothing but p-hacking: flailing around trying stuff until something works, and then claiming you knew it all along rather than found it by chance.

So those are my thoughts on the solar physics.

What about the socio-political-economic aspects?

First off: this work has been immediately politicized, apparently initially by Zharkova herself, as she has made remarks doubting the role of anthropogenic greenhouse gases in observed modern climate change.

This has resulted in equally ideological push-back based on completely nonsense claims, like "if the sun is such a key driver of the Earth’s climate, then why has the entire planet (air, oceans, land, and ice) warmed rapidly over the past 60 years while solar activity has declined?"

One can understand why anyone with two working brain cells would dismiss out of hand an argument that was based on such a ridiculous proposition.

"If weather is such a key driver of car crashes, why was there a major pile-up on a sunny day in August?"

I don't know... maybe because weather is sometimes important and other times other factors, like traffic volume matter more?

Does anyone think that pointing to a single case where a particular cause is not operative says anything whatsoever about its significance when it is operative?

That anthropogenic climate change has been the dominant factor in the past century says literally nothing about the role of the sun four hundred years ago, and anyone who suggests that this question is not just relevant but sufficient to debunk the suggestion that the sun may have a significant role to play at other times is obviously not familiar with the rudiments of effective argument.

Does the sun have a role to play? Zharkova suggests a grand solar minimum might result in as much as 1.0 C cooling over the next couple of decades. Climate modellers argue that 0.3 C is a more realistic estimate.

Either way, this is good news: at this point, any extra time we get to decarbonize is a bonus, and the way industry is moving into renewables it is perfectly possible that the extra decade we might get out of a grand solar minimum makes the difference between success and failure when it comes to saving global capitalism from itself.

Who could object to that?

So: I'm pretty doubtful about the predictive value of this model. Without actually understanding the physics behind a phenomenon, extrapolation is not possible. Remember the "cubic model" of American covid? Some idiot fit a cubic to some of the data and argued on that basis that the number of cases would become negative in a few weeks... or something like that. Since there was no underlying casual foundation for the fit, it predicted nothing.

In this case, the PCA analysis is interesting, but its decomposition and extrapolation is unwarranted. If the Maunder Minimum didn't show up in the first or maybe second try, then the result is just a fishing expedition, where they kept on trying different approaches until one got it right.

Having the underlying physics well in hand allows you to search a large parameter space to ask the question, "Are there any physically reasonable parameters that will allow this physical model to reproduce these data?" But when you start searching the space of all possible models you're on a different kind of hunt entirely, and it won't end well.

If we are headed into a grand solar minimum--which is possible--then it will buy us a little time in adapting our global economy to be carbon free. That's a good thing. We need it. But it won't "prove it was the sun all along!" or any similar ideological nonsense. It will be evidence that solar output can be a significant factor in climate, albeit (probably) a smaller one than human-generated greenhouse gasses.

On the other hand, if there is a pause or even a decline in warming over the next few decades it comes with a dire consequence, because if we don't put our house in order in the interval, then when the sun turns back up it will be on top of a much higher greenhouse baseline than today.

That means that when the grand solar minimum is done, we will see a spike in warming that will be far larger and faster than if there had been a steady growth over the intervening decades. That kind of sudden shock would be exceptionally bad for the global economy, and for the lives of those people--which is everyone, to first order--who depend on it.

Contact Home
Copyright (C) 2018-2019 TJ Radcliffe