Space.com has an interesting article about recent research by Santer et al. (2018, Science 361, 245) into how the seasonal cycle of temperature has changed. It’s a topic we’ve looked at recently for surface temperature in the USA. Santer et al. study it in the troposphere (both TLT, the lower troposphere temperature, and TMT, mid-troposphere temperature), and look for patterns over the whole globe.
Result: a human-caused signal is now evident.
They use satellite data for temperature in the troposphere, from three sources: RSS (Remote Sensing Systems), UAH (Univ. Alabama at Huntsville), and STAR (NOAA Center for Satellite Applications and Research). They also investigate how different versions of those data sets affect the results, by using both current and former releases of their data. With these, they determine the trends in annual mean temperature over space and time (referred to as TAM(x,t)), and the size of the annual cycle of temperature (TAC(x,t)). They do the same for the results from a suite of computer simulations of climate.
Then comes the heart of the study: they undertake fingerprint analysis by checking how well the patterns we’ve observed with satellites match the patterns found in computer model simulations.
The most interesting aspect is probably the geographic pattern of trends in the annual cycle. Computer models (a large multi-model ensemble) show considerable variety around the globe, illustrated thus for TMT:
The most pronounced feature is an increase in the size of the annual cycle since 1979 (when satellite observations begin) in mid-latitude regions of both hemispheres, most pronounced in the northern hemisphere (as is well shown in the diagram above). Other features include a decrease in the size of the annual cycle in the Arctic (also visible in the diagram).
The use of a multi-model ensemble tends strongly to “smooth out” the kind of fluctuations that happen from place to place; hence we can’t expect the observed pattern to match such an ensemble mean. But if the overall patterns of observations (the “big picture”) matches well enough with the overall patterns of the computer simulations, that’s powerful evidence that the models are getting the big picture right, and that features which the models say are due to human influences really are due to human influence, in the real world.
And so they do.
Here’s the observed pattern of the rate of change of the annual cycle using RSS data (red for bigger seasonal cycle, blue for smaller):
Although there’s a lot of geographic variation which the model ensemble doesn’t show, the “big picture” patterns are there and quite pronounced: the seasonal cycle has gotten bigger in the mid-latitudes of both hemispheres, more so in the northern hemisphere, while the opposite has happened in polar regions, most strongly in the Arctic. It’s a pronounced fingerprint forecast by computer models, which has now been observed and verified in actual observations. And, the models reproduce this pattern not because of natural variation, but because of human-caused climate change.
Naturally I was curious whether or not similar behavior is shown in surface temperature (as well as mid-troposphere temperature), so I used gridded data from NASA to compare the mid-latitude band from latitudes 40°N to 60°N, to the Arctic latitude band from 70°N to 90°N. I computed the trend rates for each month of the year from 1979 (when the satellite data begin) to the present by two methods: least squares regression and Theil-Sen regression. Here are the monthly trend rates (least squares in red, Theil-Sen in blue):
The size of the annual cycle is roughly the difference between July’s temperature and January’s. According to the NASA data, in the 40°N to 60°N latitude band this difference has been increasing by 0.12 °C per decade, while in the 70°N to 90°N latitude band is has been decreasing by 0.68 °C per decade. So yes, the pattern (at least in the northern hemisphere) of mid-latitude seasonal cycle increase and high-latitude seasonal cycle decrease is present in surface temperature data also.
There’s a great deal more to this research, and I’m impressed with the depth and rigor the authors apply to their study of the data. I’ll leave further exploration to interested readers, but emphasize again the overall conclusion, which is well summed up in their closing paragraph:
Across the most recent versions of observational TMT datasets, structural uncertainty in the geographical pattern of trends appears to be smaller for annual cycle amplitude than for the annual mean (Fig. 2, A to F). This is advantageous for detection and attribution studies. Furthermore, we note that the annual cycle of tropospheric temperature is not routinely used in model evaluation. It is highly unlikely, therefore, that the positive fingerprint identification results obtained here for the annual cycle could be due to model tuning. The best explanation for these results is that basic physics and basic physical mechanisms are driving the large-scale changes in TAC(x,t). For tropospheric temperature, a human-caused signal is now evident in the seasonal cycle itself.
This blog is made possible by readers like you; join others by donating at My Wee Dragon.
Open Mind 
Thanks for another useful summary! Let me guess: once again, UAH is the ‘odd man out.’
[Response: Specifically, UAHv6 is the odd man out. UAHv5 gives results more in line with both former and present versions of RSS and STAR.]
Largely unrelated, but has anyone looked at trends in average wind strength around the world? I’m unsure of the nature of wind data collected, but you’d imagine there’d be enough of it to see a trend if one is present.
Just to be certain, did Santer et al. (2018) look at how the temperature difference between the seasons of any given year increase over time, or did they just look at the general rate of warming for various regions? I admit I got a bit confused by the second and third graphs.
[Response: They did both, but their focus was on the size of the annual cycle (difference between summer and winter). My graphs are of trend rates, in order to compare them for different months, because the data I have isn’t absolute temperature but anomaly values.]
Also I noticed those enormous error bars; do you have any idea why they seem centered around the northern hemisphere’s winter months?
[Response: Not sure what you mean by “centered around the northern hemisphere’s winter months.” Winter months show larger fluctuation than summer months, hence the larger error bars during winter months.]
Fascinating. My default assumption would have been to expect smaller annual cycles globally. My back-of-the-envelope reasoning would have been that if radiative forcing from GHGs is seasonally homogeneous (a reasonable first-pass assumption), then because of the non-linear relationship between forcing and temperature (Stefan-boltzmann), warming would be fastest under the coldest conditions. This of course does not take into account snow-albedo feedbacks (I’d assume largest in late fall due to delayed snow accumulation), non-homogeneous water-vapor effects (no idea what their seasonal cycle might be), or changes in atmospheric or oceanic heat transport patterns (again, no idea about seasonality).
I should look at the paper and see what their physical intuition on this pattern is… but very neat that it is seen in both the data and the models!
“…if radiative forcing from GHGs is seasonally homogeneous…”
Pretty sure it’s not, at least not for temperate and Arctic zones, since water vapor is a GHG.
I note the denio-sphere have posted reactions to Santer et al (2018).
On the strange planet of Wattsupia, the Lord High Denier posts the nonsense of one Eric Worrell that amounts to no-more-than blowing a rasberry – well done Eric Morrell.
And the Gentlemen Who Prefer Fantasy have posted the deep thoughts of their resident guru, David ‘I-Don’t-Understand-A-Word-Of-This’ Whithouse who argues that the SAT records do not conform to the Santer et al TLT/TMT finding (Witlesshouse actually says “there seems to be some evidence” that SAT data contradicts the Santer et al findings [but gives no support for this bold assertion] and further that “most studies (of the SAT data seasonal cycle?) show no seasonal cycle effect as seen by Santer et al in the lower troposphere” [but again gives no support for this bold assertion].) So it is interesting that the OP shows the results of actual analysis on actual SAT data rather than the vaccuous assertions of Witlesshouse, thus demonsrating that the blather of Witlesshouse is yet-again naught but bullshit.
[Response: I wasn’t trying to prove anything, I was just curious whether we saw the same in surface temperature data. Lo and behold, it was there. David Whitehouse was trying to deny it. Lo and behold, no evidence given. Draw your own conclusion.]
“…there seems to be some evidence”… if you squint a bit, and hold your head just so…
What can one say? The data look different from the vantage point of one’s colon.
Yeah, there’s a adjective for that.
I am trying to muddle my way through this and would appreciate any help that can be given. My take away is the annual temperature difference is increasing because, except in the Arctic, summers are warming more rapidly than winters and this is what was and is expected to happen. Can anyone point to articles explaining why this is the expectation and why it is happening (or give a short explanation)? I thought I had a grasp on this, but I had thought winters were suppose to warm more than summers.
Eric,
The pattern of change-in-seasonal-cycle is a finding from climate models which leaves any definative explanation far from clear. I don’t think the Santer et al (2018) paper (PDF here) will help either, although its references may somewhere attempt an explanation.
Having a stab a a very simplistic explanation, consider that a warmer world will result in an increase in the energy flux into the Arctic, particularly during the boreal winter when the temperature gradient is at its maximum, thus the reduction in the seasonality in the Arctic. Yet more of the the polar cold is expected to ‘leak’ out into lower latitudes during the boreal winter, decreasing seasonality in temperate latitudes. A simplistic rationale for this extra ‘leakage’ could be the changes in the atmospheric cells under AGW, in that the Hadley Cells & the Ferrel Cells are spreading polewards and the Polar cell is weakening, allowing the jet stream to create larger loops and meanderings, these providing the points of ‘leakage’ which will be most powerful when the temperature gradient is largest, thus during the boreal winter.
Here’s another couple of thoughts to consider, not instead of Al’s suggestion, but in addition to it. In a way, it amounts to flipping the framing of the question.
First, look at the error bars on the 40N-60N plot: especially for the winter months they are enormous (probably because winter temperatures are highly variable, especially for continental climates outside the tropics.) But that means that it’s not so clear that the pattern is really ‘faster warming in summer’, though there’s a suggestion that that could be the case in the means. I won’t attempt to make any statement about statistical significance!
Then compare the Arctic plot, and you see quite a different situation. The signal much more dramatic–not only is the vertical difference between trend estimate points greater, but the scale of the Arctic plot is triple that of the temperate one–and the summer error bars are relatively small. In fact, June and July show no overlap whatever with five of the other, colder months. So it’s very clear indeed that warming in the Arctic is primarily in the winter months, with much less occurring in the summer. (Actually, comparing values on Tamino’s plots (and remembering to compensate for the different scales), the Arctic summer warming seems roughly comparable to the temperate zone warming.)
So the pattern difference seems to be Arctic-driven. And I’d suggest that there’s a good reason. Consider the polar projection below (fingers crossed for a correct HTML link).
The Arctic circle is shown; it’s at roughly 66N, so the 70N zonal boundary is poleward from there. (For a visual reference, it roughly coincides with the north coast of the Canadian mainland.) It’s pretty clear how much this area is dominated by ocean, and even the terrestrial areas are largely ice-covered for much or all of the year.
What that means in turn is that when the temperature tries to warm in the spring, a whole lot of ice has to warm up and melt before things can *really* warm. We hear a lot about the thermal inertia of the oceans, and rightfully so, but there’s also huge thermal inertia in the ice, due to the large energies involved in phase changes in water. Some commentators have even used the word “clamp” in this connection, dramatizing the effect on temperature.
Whether or not that’s over the top, a great chart to observe this in action is the DMI ‘north of 80’ timeseries.
Almost every winter has a wild-looking swing or three somewhere, while summer temps always follow the mean relatively closely. (Actually, this summer has been relatively cool on the scale of things, or we’d probably be looking at another record-low minimum in September.) Also, if you go to the actual link–and I’ll repeat it below, as I don’t think the img tag I used will result in a clickable link–then you can click back through the plots for previous years, which will also show pretty clearly how the winters have warmed on average since the beginning of this record back in 1958.
http://ocean.dmi.dk/arctic/meant80n.uk.php
Thanks for the response Al Rodger. Your link doesn’t work for me and, not being an AAAS member, I would have to pay for access to the full paper. I get the idea of ‘leakage’ from the Arctic, and understand how it causes regional, temporary cold snaps, but don’t understand how that could be responsible for lessening winter warming enough to matter. I”m just confused. Tried to find anything about this in the IPCC reports and found some figures but no discussion or explanation. From the figures, I know I’m just eyeballing, but I don’t see summers warming faster than winters.
Click to access WG1AR5_AnnexI_FINAL.pdf
Eric,
Regarding access to the full paper, I am no AAAS Member but it is possible that I am Registered with AAAS which allows me access.
While the paper is stacked full of explanations for various warming features seen in models and satelite data, the paper rather skates over such explanation for the enhancement of the seasonal cycle at mid-latitudes. It does in its Fig 2 show the mid-latitude enhancement is less pronunced in the CMIP5 model average than it is in the satelite data and in Fig 4 it shows the satelite data featuring what the paper call “small-scale differences” to the CMIP5 averages. But these apply to TMT altitudes. The paper shows that with the lower TLT altitudes, the modelled features are smaller and it reports that it remain undetectable in the TLT satelite data, although this applies to the “fingerprint” as a whole.
What we are still lacking is surface model outputs showing the mid-latitude increase in seasonal cycle. As you say, the IPCC AR5 Figs AI.4 & AI.5 show no sign of an increased seasonal cycle. Yet the SAT data does demonstrate it (as per the OP & you can also show it using the GISS maps webpage although there is much noise in the signal).
I would suggest a dig into the Santer et al (2018) references but for you that is less than helpful without sight of the full paper.
Al, I get an invalid or corrupt file-message for the Santer et al paper.
Marco,
That is now happening when I try to re-visit that Santer et al (2018) PDF. Also this page which used to give the full text for me now presents a paywall.
Doc and Al, thanks for the replies. Making more sense to me as I read more. Followed other links through the space.com article which offered some explanation.