What about the solar cycle? Yes, John V, that could explain the falsification.
I’ve trying to avoid discussing the favorite topic of all skeptic blogs. You know what I mean: The all powerful sun whose dimming may have caused Kristen Byrnes to Ponder the Maunder and a frequent commenter at Tamino’s to post comments like:
Eli Rabett // April 5, 2008 at 11:59 pm
What part of there are no 11 or 22 year cycles in the global temperature record don’t folks understand?
Well, Eli, I don’t know what part folks don’t understand.
Nevertheless, John V, who like me, believes AGW is the correct explanation for recent warming, pointed out that there are many who believe solar cycles do have a noticeable effect on the GMST. Moreover, tucked away in Chapter 9 of the AR4, is a rather brief statement suggesting that there may be a 0.1C peak to trough variability in temperature associated with the 11 year solar cycle. (Quote and link here). Chapter 10 describing the actual projections suggests the IPCC mostly omits solar forcing from the SRES. Then, on top of that, JohnV suggested that as it is a known fact that 2001 was near the peak of the solar cycle, and we are now somewhat near the bottom, this could explain the falsification.
Now, despite my aversion to stepping into the arguments about the sun’s impact of the climate, I do have a policy of considering any specifically identified cycle, with a conceret estimate for it’s strength, that might redeem the IPCC projections. (Sorry, but I can’t look at non-specific suggestions. 
)
Since JohnV suggested one, I confirmed the IPCC AR4 does, indeed, mention the possibility that solar cycles matter and with quite a bit of moaning and wailing, I grudgingly agreed to consider the solar cycle. (JohnV will confirm my grudging behavior. I was much more positive about ENSO, because those advocating AGW always discuss ENSO. In contrast, mostly spectics or denialists seem to mention the solar cycle.)
The results? If we assume the 11 year solar cycle imposes a perfectly sinusoidal influence on the Global Mean Surface Temperature, an amplitude of 0.05C, and with a peak centered on Jan 2008, then the IPCC results no longer falsify under this assumption.
Mind you, the central tendency of the measured trend is only 0.8C/ century, which is quite a bit lower than 2C/century, but given the range is -1.4 C/century to 3 C/century, so 2 C/century drops in line.
Here is the graph, with uncertainty intervals. The brown line associated with 2C/century falls between the two fuzzy orange lines. (Note that 0C/century also falls inside the fuzzy orange lines.)
Answers to Obvious Questions
- Did I do anything really fancy? No. This is quick and dirty. You can believe it if you like, or not. With luck, we’ll get discussion and people will dredge up references about the solar cycle and its influence on global mean surface temperatures so I won’t have to!
- Why 11 years? This is the periods “everyone” pretty much believes is associated with solar cycles. “Everyone” includes NASA GISS.
- Why an amplitude of 0.05C? Chapter 8 or the technical summary of the AR4 suggests a peak-to-trough change of 0.1C (i.e. 2 * 0.05C) is suggested for the 11 year solar cycle.
- Why set the peak of the effect at Jan 2001? This is a combination of the known time of the peak and cherry picking. The peak in the total solar irradiance may be earlier. However, based on phenomenological arguments, there could be a lag of up to 90 degrees in the response. (All things being equal, longer lags will generally be associated with in smaller measureable responses.) I picked the peak to corresponds to a time that improved the linear fit from 1979- now as best as possible. (For what it’s worth 0.05C also improves the fit more than the Camp and Tung suggestion of roughly 0.1C for the solar effect.)
- What do climate models say about the effect of solar variations on GMST? I don’t know.
Steve mosher suggested we examine the data from NASA GISS MODEL E. Evidently, we can examine the averaged results of 5 ensembles. John V downloaded the data, ran an FFT and the model seems to indicate the effect of the 11 year solar forcing is negligible. FFT should reveal strongly cyclic data, particularly in an ensemble average where weather noise is reduced. However, Steve, John and I aren’t entirely positive we are interpreting the NASA resource properly.
- Does this mean 2C/century is confirmed? No. There are two reasons for this.
First, even if you embrace the idea that solar cycles have a measurable effect on the GMST, and the recent flat trend is due to the dip into the bottom of the solar cycle, the 2C/centruy is not “confirmed”. It is only “not falsified”. This is because the “type 2″ (’Beta’ or β ) error for this test relative to 0C/century is 49%. That means that even if 0 C/century is correct, there is only a 49% chance of “falsifying” 2C/century. The range obtained using the simple minded solar model is -1.4C/century to 3C/century.
In contrast, if the entire idea of the solar cycle having a discernible effect on GMST makes your eyes roll, the 2C/century should, in your mind, falsified by the current data with no specific culprit identified. There is still the PDO or AMO, and if someone wants to suggest literature explaining the probable magnitude of their influence, I’m always game to look.
- Do I, Lucia, believe this is “the” explanation for the post 2001 flat trend? Beats me! Will mainstream climatologists say “Oh yeah, the solar cycle is what I personally meant when we alluded vaguely to hypothetical, unnamed cycles or waiting a long time?” Beats me!
But the solar cycle does happen to start at the right time, and published estimates of the strength of the 11 year cycle are of the correct magnitude.
I am of the opinion that if climatologists believe the fairly regular, fairly predictable 11 year solar cycle really has an 0.1 C effect, they should be incorporating this effect into their projections and discussing its probable magnitude in very explicit terms and in discuss it in prominent places in documents written for public dissemination.
For now: I’ll be carrying the “solar correction”, along with a “without solar” comparison forward. As we move into the future, the sun’s intensity is expected to increase, and we’ll see whether warming tracks the expected intensity rebound in real time.
Future work
Some will note I didn’t fold in ENSO.
ENSO explains a lot of the variability, and so adding it explicitly tends to tighten the uncertainty intervals, and potentially reduce β error. However, every time I start getting back to trying to figure out how to deal with that better, someone suggests something else that might be useful to look at sooner. (This can be real work dealing with cyanide canisters or the question of the solar connection.)
So, I am planning to figure out how to incorporate ENSO in some not too idiotic way next month. As I said, I’ll carry both “solar corrected” and “not solar corrected” forward. That way, we can all see whether or not the solar correction ends up tracking temperatures as the solar intensity turns up. (Given the prominence of this in climate-blog-wars, people might be interested anyway.)
But meanwhile: If you believe the sun matters, the IPCC projections may yet redeem themselves. (Though you can wonder why the panel doesn’t include this effect in their projections. Sorry, but the fact that it averages out doesn’t cut it with me.) If you believe there is no 11 year solar effect, then they projections aren’t looking good.
Concrete suggestions invited; (the right to bitch and moan while considering them reserved.)
Previous Post:
« UAH vs RSS Temperature Differences: Why energy at 1 year?
Next Post:
Stratospheric Temperatures: Bleg. »
54 Responses to “What about the solar cycle? Yes, John V, that could explain the falsification.”
You can leave a response, or trackback from your own site.
John V April 24th, 2008 at 11:22 am
I’m impressed that you wrote this up.
I was beginning to doubt your neutrality but you have been redeemed.
(Let the flames directed at my neutrality begin).
One question about your answer to obvious question #3:
If IPCC says 0.1C peak-to-peak, and the trend from Jan2001 to Jan2008 is from peak-to-peak, why use only 0.05C?
lucia April 24th, 2008 at 11:40 am
Hhnmmm I think it’s peak to trough… I have the quote:
The 0.05 C is center to peak. So, it goes -0.05C at the bottom to +0.05C at the top, which is 0.1C peak to trough. That’s the way I interpreted what is actually shown in a tiny little graph in “H Gleisner, P Thejll - Geophys. Res. Lett, 2003 - dmi.dk” which is cited in Chapter 8, of the AR4 and available in PDF. I considered that paper sort of “enough” to justify where the 0.1C might come from.
Let me go fix that if I wrote peak-to-peak!
Also, since we have that Camp&Tung reference. I fiddled a bit. See the “cherry picking” admission? I took the 1979- now data, and tested which level results in the most “improvement” of the fit for that, measured by the best R value by varying the location of the ‘peak’ and the magnitude of the amplitude.
I admit I groused– but I do always feel the need to read a bit to see if shoving in a “fix” really makes sense. I’m still not sure the climatologists totally or mostly agree that there is a 0.1 C peak-to-trough. I still think that if they think there is a ±0.1C effect that is as predictable as the sun, they ought to mention this quite prominently and very near the actual projections.
No matter what, they will be re-evaluating before 2014, and like it or not, it’s best to consider data that really, honestly came out after predictions were made. Even with 14 years, a ±0.1C for with an 11 year period is important especially if you use “slide and eyeball”, as done in Rahmstorf, or the validations portions of the IPCC.
John V April 24th, 2008 at 12:10 pm
Sorry, I wasn’t very clear. I meant peak-to-trough. I agree with your assessment of plus or minus 0.05C…
…and now I see my mistake. I thought you were applying a correction for a solar influence of -0.05C from Jan2001 to Jan2008. I now understand that you are using a circular solar cycle with an amplitude of 0.05C. Sorry for my confusion.
I’ve been toying with adding a simulated solar cycle GMST response to GIS temperatures. I took the solar forcing from the data file you used for Lumpy, subtracted a running 11-year mean, and scaled it to temperature with a peak-to-trough of 0.1C. I ran a couple of analyses with this simulated temperature response added and subtracted from the GIS temperatures. The FFT amplitudes at the 11-year period were:
Raw GIS: 0.022 degC
GIS + Sim: 0.040 degC
GIS - Sim: 0.01 degC
There are a few things to note from this:
1. There is something wrong with the scale;
2. The response nearly doubles when adding 0.1C solar response;
3. The response is nearly eliminated when removing 0.1C solar response
This calls into question my FFT analysis of the Model E response to solar forcing. It *may* suggest that the raw GIS temperatures show approximately 0.1C peak-to-trough for the solar cycle.
lucia April 24th, 2008 at 12:17 pm
I don’t have any FFT codes handy, so… But maybe someone will run it. I ran across interviews were various climatologists claim the AOGCMs do reproduce 11 year long variability, and others that say they don’t. So, I don’t know what’s claimed.
But in the table in chapter 10 of the scientific basis for the AR4, the GISS models listed “Y” and “Y” for accounting for solar. If AOGCM showed no response, I’d be surprised they would bother to integrate in the SRES. I’m saying this because… well… I wouldn’t go to the effort to add complexity to predict no response. Other models listed “C” and “C”.
John V April 24th, 2008 at 12:34 pm
Ok, I’ve played with my simulated temperatures on top of GISTEMP temperatures a little more. To me, they suggest an 11-year cycle with a peak-to-trough amplitude of ~0.13C.
When I subtract a 0.13C temperature response to the solar cycle (with the exact shape of the actual solar forcing), the FFT at ~11Hz drops to near zero. When I add the same temperature response, the FFT at ~11Hz doubles. I think the assumption of a circular solar cycle makes the response hard to find. It is easier to find using the actual shape of the solar cycle (faster rise, slower decline).
I concede that my method may be completely invalid. Anyone interested in testing?
lucia April 24th, 2008 at 12:40 pm
John– do you have literature on the actual shape? Or do you calculate it based on the published TSI?
Initially, I was thinking of looking at the published TSI, but then I thought I’d do quick and dirty first. (I’m not sure one can do much better than quick and dirty with this.)
Alan S. Blue April 24th, 2008 at 12:44 pm
One of the issues with comparing “with an 11-year cycle” is that the sunspot cycle is, in itself, weather. Solar weather. We’re currently in our thirteenth year from the last sunspot minima, with not much sign of the next cycle beginning. Then you add potential lags, and the uncertainty in what precisely should be the significant solar parameter.
I’d be interested in this question:
If you compare just the years of the sunspot maximas, what is their distribution compared to the all-years-trend? If there’s an unknown-but-real contribution for which sunspots are a reasonable proxy, one would expect ‘the maxima’, or perhaps ‘the year after the maxima’ to have a decidedly non-gaussian distribution across the trend.
IOW: Before saying “Do I see an 11-year periodicity?” ask the blank-slate empirical question “Is this factor of any interest?”
John V April 24th, 2008 at 12:52 pm
lucia,
I did not calculate the shape. I just used the published TSI, de-trended, scaled it to +/-0.05, and added or subtracted it from GISTEMP historical temperatures. I then used Excel’s FFT on the last 64 years of data.
I would prefer to use monthly data but yearly was on my hard drive. I also need to figure out Excel’s FFT scaling or finally make the switch to R.
lucia April 24th, 2008 at 12:59 pm
Alan Blue–
There are lots of interesting question. My main one is: The IPCC accounted for certain things in their projections. But, evidently not others. Does the flat trend seem to be due to the things they didn’t account for in their projections.
It is interesting that the sun spots are not materializing on time. I’m not going to be looking into the sunspot- sun intensity link though. I think plenty of people have that covered!
Hans Erren (7 comments.) April 24th, 2008 at 1:00 pm
Lucia are you familiar with this multivariate analysis?
Douglass, D.H. and B.D Clader, 2002, Climate sensitivity of the earth to solar irradiance,
Geophys. Res Lett. vol 29, no. 16, 10.1029/2002GL015345
lucia April 24th, 2008 at 1:02 pm
Hans– I haven’t read that. I know Gavin and someone were arguing about it over on a thread that Brigg’s blog.
Hans Erren (7 comments.) April 24th, 2008 at 1:13 pm
here is figure 2 of DC2002
Robert Beck April 24th, 2008 at 7:34 pm
I did an FFT analysis of monthly sunspot data (1850 through 1995) from NGDC, and eliminated all harmonics except those required to produce a wave-form that closely matched that of the smoothed envelope of monthly sunspot numbers. Those harmonics ranged from approx 7 to approx 15 year periods. I then ran an FFT analysis of monthly global temperature anomalies (1850 through 2008) HADCRUT3, eliminating all harmonics except the those used to generate the sunspot-envelope waveform. Interestingly, the resultant temperature wave-form was quite similar to the sunspot waveform, and the peak-to-peak temperature was very nearly 0.1 deg C on average. However, phase-relationships between the two waveforms were all over the place, ranging from in-phase to 180 degrees out-of-phase and back, with no discernible pattern.
From this analysis, any relationship of temperature to sunspot cycles seems suspect, and it is questionable whether the 11-year component of the temperature signal can be definitively ascribed to solar cycles, at least with simple mathematical/statistical procedures. Is it possible that there is a mechanism that transfers sunspot-related energy to the atmosphere with varying amounts of lag?
Since the sunspot cycle periods vary significantly in length, no single sine-wave can adequately describe the wave-form, however 6-8 harmonics, with periods ranging from 7 to 15 years, will reproduce the smoothed data very well. I also used Excel’s FFT routines, as I am not yet well-acquainted with R.
I did de-trend the temperature data, and zero-padded both sets of data to eliminate end-effects and to raise the number of input data to 2048.
Raven April 24th, 2008 at 8:06 pm
Lucia,
I don’t think it makes sense to include ENSO correction and the solar correction in the same analysis since they are are likely one in the same (i.e. solar maxes correspond to El Ninos and solar mins correspond to La Ninas). This could mean that the solar correlation is accidental or their is an unknown link between the solar cycle and ENSO. In either case, using the solar effect to avoid falsifying the 4AR predictions does not actually help the IPCC cause because they cannot explain why the solar cycle has the effect claimed.
Bob Tisdale April 25th, 2008 at 2:02 am
Raven: Regarding your statement that El Ninos correspond with Solar Maximums and La Ninas with Minimums, I find no correlation between MEI and TSI.
http://i26.tinypic.com/292lisg.jpg
The MEI is annual data and I subtracted 1366 from the TSI to scale it.
lucia April 25th, 2008 at 4:19 am
Raven:
I agree that adding Solar doesn’t necessarily help the IPCC case. But, it is useful to see if any particular mechanism with longer periods, and that is omitted from the projections, and that particular one, suggested by JohnV, happens to do it. If climatologists were to say they believe that effect is there, and fairly well know to follow the sun’s cycle, I would suggest they should then be incorporating that into the projections. Given the range of warming predicted, and the true time frames for assessing validity (which, no matter what they say, is from the publication of one report to the time they start writing the next) it’s to big to just ignore.
I’d been looking at AMO, and at least from what I read, while it’s a long period, that switch can’t be the cause. The other issue is PDO– that switched around the right time. But I haven’t found any estimates of how much temperature variation that might cause from peak to trough.
Oddly, I looked at the standard deviation of temperature for the GISS “solar” runs, and they aren’t far off from the ones we see in recent data. That’s interesting because the GISS solar runs don’t have volcanic activity. The match sort of implies this time period is long enough to get roughly correct error bars, as those do scale with the standard deviation. (I need to look a bit more and consider the “measurement noise”, but they really aren’t far apart.)
lucia April 25th, 2008 at 4:22 am
Bob–
I think Judith Lean suggested solar may trigger something but not necessarily ENSO. I need to collect some of these papers together, but it appears at least some thing the solar connection may not be the actually magnitude of TSI, but an indirect response. ENSO cycles are shorter than 11 years.
steven mosher April 25th, 2008 at 4:35 am
lucia I checked the MITGCM code for incorporation of an 11 year cycle, I couldnt find any. Next I’ll check modelE
i started with MIT because its better documented, with modelE i’ll have to slog through code, or write gavin.
the former being more fun
PaulM April 25th, 2008 at 4:55 am
Those rabbits and others who claim there is no evidence of the solar cycle in the temperature record should try looking at some data. Look at HADCRUT3 and you see peaks at roughly 1940, 1950, 1960, 1970, 1980, 1990. Now look at the sunspot data for example here and what do you see? Peaks around 1940, 1950, 1960, 1970, 1980, 1990.
Note that in the late 20th century the solar cycle length is about 10.5 years. The cycle tends to be shorter when the amplitude is large and longer when it is weaker.
anonymous April 25th, 2008 at 5:41 am
I thought the premise (AGW) was that the sun had NO effect on climate (AGW) but now it seems it may. after all…LOL
Raven April 25th, 2008 at 6:07 am
Bob Tisdale says:
The match is not perfect because there are additional wiggles but the pattern is consistent: as the cycle falls from max to min the mei goes from high to generally low (with blips). The reverse happens on the rising edge. There is enough of a pattern to say that any observed correlation between solar cycles and temps could be simply be picking out one of the patterns in the MEI.
steven mosher April 25th, 2008 at 7:37 am
as long as the sun nuts dont dominate or derail the discussion, I’m ok with it.
I’d boil it down to one question. When the GCM contributing to the IPCC make a “forecast”
do they assume a constant solar forcing or varying forcing. If varying, then how
is this implemented.
Is that the right question? JohnV care to weigh it, before I shoot an email to gavin
John V April 25th, 2008 at 7:46 am
anonymous:
The premise of AGW is *not* that the sun has no effect on climate. That’s the strawman that some use to confuse things. The IPCC position is that many things affect climate, including greenhouse gases. It is the position that greenhouse gases have no effect.
Raven:
Do you have any references or analysis for your purported relationship between ENSO and the solar cycle?
lucia:
As you know, the IPCC assessment reports are basically literature reviews. If we want to know the “opinion” of the IPCC regarding the solar cycle, then we should use the references cited in the IPCC AR4 WG1 9.2.2 (listed below). The text of AR4 is pretty clear about 0.1C peak-to-trough and there is no mention of any controversy. Tamino and Eli seem to disagree but have not published anything.
The solar cycle effect on trends is significant for this analysis only because of the unfortunate coincidence with the solar cycle.
—
IPCC AR4 Solar Cycle References:
van Loon and Shea, 2000:
van Loon, H., and D.J. Shea, 2000: The global 11-year solar signal in July-August. Geophys. Res. Lett., 27, 2965–2968
Gleisner and Thejll, 2003:
Gleisner, H., and P. Thejll, 2003: Patterns of tropospheric response to solar variability. Geophys. Res. Lett., 30, 44–47.
Haigh, 2003:
Haigh, J.D., 2003: The effects of solar variability on the Earth’s climate. Philos. Trans. R. Soc. London Ser. A, 361, 95–111.
White et al., 2003:
White, W.B., M.D. Dettinger, and D.R. Cayan, 2003: Sources of global warming of the upper ocean on decadal period scales. J. Geophys. Res., 108, 3248, doi:10.1029/2002JC001396.
Coughlin and Tung, 2004:
Coughlin, K., and K.K. Tung, 2004: Eleven-year solar cycle signal throughout the lower atmosphere. J. Geophys. Res., 109, D21105, doi:10.1029/2004JD004873.
Labitzke, 2004:
Labitzke, K., 2004: On the signal of the 11-year sunspot cycle in the stratosphere and its modulation by the quasi, biennial oscillation. J. Atmos. Solar Terr. Phys., 66, 1151–1157.
Crooks and Gray, 2005:
Crooks, S.A., and L.J. Gray, 2005: Characterization of the 11-year solar signal using a multiple regression analysis of the ERA-40 dataset. J. Clim., 18(7), 996–1015.
John V April 25th, 2008 at 7:50 am
steven mosher:
I would expect that if the solar cycle influence is fed into the models, it would be through a data file of TSI. I really hope there’s no 11-year cycle in the code — that would be sloppy even for research code. Your question seems like the right one and it can’t hurt to ask. He may only be able to answer for NASA’s models.
I think lucia found a table that shows which models include the solar cycle (something about “Y” and “C”). If I understood her correctly, only some models consider the solar cycle.
I’m going to have a look at the Model E response again and try to figure out how Excel scales FFT results.