Sunspots and snow

Bear with me here. I come from a place with a long history of nutty sunspot-based weather prediction, so I’m well aware of the pitfalls. But our sun really is a very slightly variable star, and aspects of our climate do seem to very slightly vary with its output — particularly snow cover. A couple of years back Ken Green at NSW NPWS co-authored a paper¹ showing a modest local effect, which I’ve been meaning to look into.

Solar output correlates with the number of little black spots on its surface, called sunspots². People have been counting them for hundreds of years:


You can see the well known 11 year cycle, and that there is additional, longer duration variation superimposed on that. The record since our best snow depth series began in the 1950s looks like this:


It’s usual to number the cycles as shown. The most recent two have been atypical. Cycle 23 was slow to complete, taking nearly 13 years, and cycle 24 has been weak as well as very late. There is variation on top of the variation, and none of it is truly regular.

Good measurements of total solar output are only available for the satellite era. The correlation looks like this:

Total solar irradiance and sunspot number, from NOAA via climate4you

Despite sunspots being black areas that radiate little, total solar output increases when the number of sunspots increases. Sunspots are huge magnetic disturbances on the surface of the sun, and their presence indicates increased overall activity. Look at the red trace against the left hand scale. The average change in solar output from the trough to the peak of the cycle is tiny — about 1 W/m² in a total output of nearly 1400 W/m² (at Earth orbit). That’s about 0.07%. Nothing, right … except when you consider that the radiative forcing due to doubling of the atmospheric concentration of carbon dioxide is only about 3.7 W/m², and that’s expected to change the average surface temperature by about 3°C, eventually³. The solar effect is small, but not quite negligible. One reason we don’t see larger temperature swings from it may be that it happens too rapidly. The climate system has large thermal inertia (mostly from the oceans), and doesn’t respond much to the “rapid” signal of an 11 year cycle.


Snow depth correlation

Sánchez-Bayo and Green (2013) report relatively strong correlation between Spencers Creek snow depths (midway between Perisher Valley and Thredbo, NSW, from Snowy Hydro) and the average half-cycle sunspot counts for the 12 available half cycles of data (R² ~ 0.5, p ~ 0.01). There was also a suspicion of an 11 year cycle in my Fourier analysis of the Spencers Creek peak depth record, though as I pointed out at the time, it’s well down in the noise:

Spencers Creek peak snow depth cycles

Spencers Creek peak snow depth cycles

It’s much more useful for me to look directly at correlations with peak annual Spencers Creek depth, which I’ve of course used extensively for my snow depth prediction model. For that I’m just going to try a bog-simple 12 month trailing average sunspot number to the end of August (the end of winter). So here goes; it comes out like this:

Spencers Creek season peak snow depth and sunspot number

Spencers Creek season peak snow depth vs sunspot number

Well that’s not very exciting. The correlation coefficient is only 0.05, so sunspot number “explains” about 5% of the variance in season maximum snow depth. That’s about the same amount as Pacific Decadal Oscillation and only a quarter as much as sea surface temperature. Nevertheless, provided the correlation is real, robust, and independent of the others, it has the potential to make another small improvement to my snow depth prediction model.

But there’s just one little niggle here. Both my correlation and Sánchez-Bayo and Green’s show snow depth increasing with sunspot number. Intuitively the correlation should be the other way: sunspots -> more solar output -> hotter -> less snow. Without at least the gist of a physical explanation, using correlation for prediction is fraught.


The standard explanation for inverse sunspot correlations (also observed, for example, in the flow of some South American rivers) is that other effects of the solar cycle alter upper atmosphere chemistry and physics, leading to the production of ultra fine particles that settle into the troposphere, where they function as cloud condensation nuclei (CCN). More clouds lead to more rain (those rivers) and more snow. The postulated effects are from increased solar ultraviolet output, affecting upper atmosphere chemistry, and from stronger solar wind — the continuous flow of charged particles emanating from the sun that bathes the whole solar system. Increased solar wind near the solar maximum displaces the Earth’s magnetosphere (observed), allowing increased penetration of cosmic rays (likely), which generate particle streams from upper atmosphere nuclear interactions (demonstrated in the lab at CERN). The trouble with the latter is that those particles are at particle physics scale — many orders of magnitude too small to function as CCN — and no one has managed to explain how they might coalesce to a suitable size. See the discussion at RealClimate.

Frankly, that all seems pretty tenuous to me. Note that the observed correlations are necessarily cherry-picks — we only remark on the ones we see; we don’t remark on the very many more that we might have seen but don’t. Sorry, I don’t buy it; at least not at the current state of understanding.



1. Sánchez-Bayo, Francisco, and Ken Green. “Australian snowpack disappearing under the influence of global warming and solar activity.” Arctic, Antarctic, and Alpine Research 45.1 (2013): 107-118.

This paper has multiple issues — mostly to do with dubious correlations with atmospheric carbon dioxide concentration — but it has considerable strengths, too. (Carbon dioxide concentration is a rate effect; it affects radiation imbalance. Current climate change is largely an aggregate effect; it depends mostly on the time integral of radiation imbalance, due to the large thermal inertia of the system. The two need not be closely correlated on an annual basis.)

2. Sunspots are occasionally visible to the naked eye at sunrise and sunset, but I’m not recommending you try looking. It genuinely is dangerous to eyesight.

3. This comparison is not really valid. The effect of solar output change is reduced by the planetary albedo (reflectiveness — about 0.3), and the CO₂ forcing is averaged over the whole surface of the planet, which is four times the area of the disk receiving solar radiation. A fairer comparison would be about 0.2 W/m² effective sunspot cycle solar vs 3.7 W/m² for 2xCO₂, both averaged over the whole surface.