Guest Post by Willis Eschenbach
Over in the Twitterverse, where I appear as @WEschenbach, in a discussion someone made the following comment:
Solar minimum affects periodicity of volcanic activity.
Year ‘without a summer’, 1816, and several other big volcanos occurred during solar minimum cycles.
I said that wasn’t true, and pointed him to my analysis of volcanoes and sunspots.
Someone else commented:
But interestingly, it is true for earthquakes, for which we have more occurrences. They are less likely at low sunspot numbers, because of less EM tidal force on plates. As spots clime up above ~90, their EM force evidently rises enough to help nudge the plates like moon does.
I thought … “electromagnetic tidal force”?!?
So I decided to take a look. I started by going to the US Geological Survey website to get all of the earthquake records. As is my habit, I first wanted an overview of all of the data. I’ve found through bitter experience that omitting this step can lead to bad outcomes. So I began by plotting up the number of earthquakes per year, shown in Figure 1 below:
Figure 1. Earthquake counts by year, 1900 – 2007
Yikes! You can see why I take an overview of the whole dataset before doing any analysis. Starting in 1964 they got a lot more quakes. I figured that this was an artifact of the detection and recording process, and that there would be many more small quakes in the more recent data. As seen in Figure 2, that turned out to be true.
Figure 2. Annual average earthquake magnitude, 1900 – 2007
Again … yikes.
To get around this, I decided to just look at strong earthquakes, those with a magnitude over 6.5. These were spread out relatively evenly over the whole period, while including smaller magnitude quakes biased the record to more recent times.
To compare these larger earthquakes with sunspots, I first made a “histogram” showing the percentage of days with 1-20 sunspots, 20-40 sunspots, 40-60 sunspots, and so on. This is my baseline. If the earthquakes fall at random with no effect from the sunspots, then the percentage of earthquake days in the histogram “bins” containing e.g. 0-20, 20-40, or 40-60 sunspots should be very close to the baseline histogram of sunspot numbers from all days in the same bins in the period of record. Figure 3 shows that graph.
Figure 3. A histogram of the percentage of daily sunspot numbers on the days when the 3,604 largest earthquakes from 1900 to 2007 occurred (red), compared to the histogram of the percentage of all daily sunspot levels from 1900 to 2007 (blue). The vertical axis shows the percentage of the total number of trials in each bin.
As you can see, the histogram of the sunspots on days when there are earthquakes (vertical red rectangles with red hatching) is very close to the histogram of the sunspots on all days (vertical blue rectangles).
How close are they? Well, we can determine that with a “binomial test”. It gives the 95% confidence interval (95%CI) for the percentage of quakes in each bin. The black/white vertical “whiskers” at the top of each earthquake percentage (red rectangle) show the range that 95% of random trials would fall inside if the earthquake days actually occurred at random.
And as long as the 95%CI overlaps the blue rectangles showing the expected percentage from all the days, we can say that we have no statistically significant evidence that the earthquakes actually are not falling at random. It’s worth noting that some 5% of random earthquakes will fall outside that range … and in fact, in the bin from 120 to 140 sunspots per day, the 95% CI doesn’t quite include the top of the blue rectangle. Finding one or two results like that in a sample this size is not a surprise, as we expect some 5% of random occurrences to be outside the 95% CI.
So there you have it … sunspots do NOT affect earthquakes in any statistically significant fashion. One more interesting conclusion from this—every ten days or so there’s an earthquake stronger than 6.5 somewhere on the planet. Standin’ on shaky ground …
Now in the past I’ve posted up a number of my studies of the purported effects of sunspots on surface weather variables like rainfall, river flows, clouds, sea levels, tides, and the like. There is a list of my investigations here.
And when I began my search, I was a true believer in the solar-weather connections. I’d heard the claim from the British astronomer William Herschel in 1801 that sunspots affected the wheat harvests in the UK. I thought it would be easy to find evidence that sunspots, or some other sunspot-related phenomenon like solar wind or cosmic rays or heliomagnetism or the like, would be affecting some surface weather phenomenon.
Nor, to my great surprise, I have ever been able to find one single surface weather phenomenon which shows the slightest effect from some sunspot-related phenomenon. And today’s study just adds to a long list of failures to find such a correlation.
Now, let’s be clear—it’s very hard to disprove a negative. Finding one black sheep will disprove a positive statement that all sheep are white. But not finding a black sheep doesn’t disprove a negative statement that no sheep are black … maybe you just haven’t looked in the right place.
As a result, I haven’t shown that sunspot-related phenomenon are not affecting some surface weather phenomenon. Maybe I just haven’t looked in the right place. I’d love to actually find something. At this point, it would be a surprise and a coup. All I can say is, I’ve looked at a whole heap of claimed correlations and found nothing.
So to continue my search, let me make the offer that I’ve made several times before. I invite people to send me two links—one to the one very best study that you know of showing a correlation between a sunspot related phenomenon and surface weather of some kind, and a second link to the data used in that study. If you send me those two links I’ll see what I can find, and as my mom used to say, “God willing and the creeks don’t rise”, I’ll report back the results good or bad. A few caveats:
• NO REANALYSIS “DATA”! It’s not data as we commonly understand the term. Instead it’s the output of a computer model … and computer models tend to be linear, with whatever you put in as input coming out as output. Might be lagged and transformed, but what goes in comes out. See my post here for an example. This means that since total solar irradiation is used as an input, it will very likely appear in the output … and that will mean absolutely nothing. So no reanalysis “data”, please.
• I’m asking for two links, one to the study you think is best, and the other to the data used in the study. Without the second link, I cannot even attempt to replicate the study.
• Surface weather-related datasets only, please, no atmospheric datasets. I’m a ham radio operator, H44WE, and I’ve long been aware that changes in sunspot levels are mirrored in changes in the ionosphere. But that’s from about 60 km (37 mi) to 1,000 km (620 mi) altitude, far above us, and those changes don’t seem to make it to the surface.
• Finally, please first check all the posts at the link I gave above to my previous work on sunspots. No good you go to all the trouble of digging out a link to your best study and another one to the data, only to find out that I’ve analyzed it already.
I make this request for links for just the one single best study you know of because the number of bad studies, and in some cases ludicrously bad studies, on this question is huge. So please, don’t send me some laundry list of “27 NEW REALLY AWESOME STUDIES SHOWING IT’S THE SUN, STUPID!” My life is too short to dig through piles of trash looking for a diamond. Send me two links to your best study and data, and time permitting, I’ll take a look.
My very best to everyone on this rainy slow day,
PS: Misunderstandings are the bane of the intarwebs. So I politely request that when you comment you quote the exact words that you are discussing. That way, we can all be clear on exactly what and who you are referring to.