Leif Svalgaard claims “TSI has not fallen since 2003”. It’s technically true in a sense, but demonstrably false when discussing 11 year smoothed trends (which is written on the graph he was criticizing). Willis Eschenbach sadly was carried along. This post is in response to an overheated thread at WUWT. Both men owe David Evans an apology.
The fuss is over the big fall in TSI. Leif Svalgaard said it was “almost fraudulent” that we claimed there was a fall in TSI since 2003 since there wasn’t a fall in this dataset. He says: “There is no such drop.” I say, look at the graph below, it’s even in your own data. Svalgaard provided the link to his TSI set, and we’ve included that line in the graph below. It’s the light-purple line. (Has he paid attention for the last ten years?)
In his rush to call it “totally wrong” and to declare “the model is already falsified” he didn’t notice we were talking about a trend in 11 year smoothed TSI, and the fall is evident in whole cycles (but takes some wisdom to find in daily or monthly data). I guess that’s a mistake that could happen to anyone — but some of us might ask politely before we started calling “fraud”, and saying things like “Mr Evans assertion is false [and I maintain seems to be agenda driven…” Likewise, Willis Eschenbach unskeptically follows: “as Leif points out, he’s using a bogus set of TSI data.” If skeptics toss out careless accusations, it rather cheapens the real ones.
Obviously the 11-year smoothed effect is news to Svalgaard, perhaps it’s news to a lot of people. It’s something David found because his Fourier work suggested a notch, and the Solar Model that was made with a notch filter predicted a big fall to come. From that David inferred there must have been a corresponding large drop in TSI and then he created the 11 year smoothed graph and found it (in response, it must be said, to an email from Lubos in April asking if there was an easier way to see there was going to be a big fall in temperature than through the model output).
The comments at WattsUp has been unseemly, and entirely unnecessary. (I’m sure it doesn’t help Anthony.) We will deal with other misunderstandings from the same thread (yes there were more) in a future post. The uninformed ad homs are a waste of time. What happened to common courtesy?
Compare the major datasets of TSI or proxies:
As to whether the SORCE data should have been used in the Notch-Delay Solar Model — it’s rather trivially clear that since it starts in 2003 it’s not very useful for 11 year smoothed graphs, because there is only a single point of 11-year-smoothed data. It’s no use for finding the model parameters, because the delay of about 11 years means it cannot be used to check predicted temperatures against observed temperatures yet. And SORCE might be wonderful but it isn’t useful for Fourier analysis of long term climate cycles either (it’s hard to find an 11 year delay in only 11 years of data).
Strangely too, for a commenter who I hear is familiar with solar data, Svalgaard seems to forget that the last peak of solar cycles was 2001-2002, which is not visible in the graph he linked to (SORCE wasn’t operating then). Svalgaard compares data that starts after the peak with the next peak and says “they are the same” as if it means something. It’s a tad misleading (to be polite). I’m sure he didn’t mean it that way.
The graph below pretty clearly shows how TSI from the 2003 to 2012 fits — at least in the larger PMOD scheme of things (SORCE data only covers this short era). Yes, it’s technically accurate to say that TSI now is the same as 2003. Svalgaard declares ” If anything TSI is now higher than it were in 2003.” But it is obvious that the peak of the latest cycle is a lot less than previous ones.
Svalgaard thinks science is a bloodsport
Svalgaard emailed me this morning saying “science is a bloodsport”.
I replied that it “doesn’t have to be… You could use logic and reasoning instead.”
All the facts could be uncovered faster by honest enquiring minds without malice. People who brought preconceived assumptions about “motivations” and bad-will into a science debate failed to read what was put before them. We knew David’s work was going to be difficult, and that’s why we’ve released it bit by bit. They aren’t the only ones who have not read carefully enough.
Svalgaard admits reconstructions are “guesses”
Noteworthy is Svalgaard’s honesty about reconstructions. Commenter Brad, here asked why Leif used the term “TSI-guess” in his file label, Leif responded saying: “All so-called ‘reconstructions’ of TSI are Guesses. Most of them bad. The TSI-Guess.xls file is my guess.”
TSI varies because the magnetic field of the Sun varies, and the field varies as the Sunspot Number does, so the variation of TSI is essentially that of the sunspot number [which is known] on top of a fixed background that does not vary. Based on the past decade of SORCE/TIM precise measurements of TSI, we can calibrate the sunspot number in terms of TSI. That gives me a Guess, which will not be correct in details, but will capture the gross features of the variation.
A brief history of TSI datasets
David Evans looked at all the major TSI datasets he could find in 2013, and downloaded the data available on 8 Aug 2013 when he froze the data used by the project. It’s harder to measure the sunlight than you might think, because there is a wide spectrum of light from UV to infra red. Everyone thought there was essentially no variation to measure up ’til late 1978 when people started observing it with satellites — it’s telling that TSI used to be called “the solar constant”. But there is really only one observed record that runs through the last 35 years, namely PMOD. ACRIM provides data from 1978, but before 1992 or so its results disagreed with PMOD and Lean (which is a reconstruction guided by PMOD). Like everything in climate, there is a war going over the adjustments and reconstructions and no one can agree. Fans of the IPCC now say the TSI was falling for decades, while others think TSI pretty much stayed high til the 23rd cycle and the 24th has been strikingly small.*
The PMOD dataset is the longest running continuous TSI record. ACRIM data got compromised by the Challenger Shuttle exploding, and then its results apparently didn’t quite make sense until the early 1990s. Bring on SOURCE, a new hot tool in 2003, which seems to work well. Lean and co used the PMOD data with sunspots during the same period and figured out a kind of calibration to use so we could estimate TSI from the old sunspot data. Yes, it’s difficult and we all wish Napoleon had fixed his satellite program, but it was not to be. The data is what it is. We are all doing the best we can.
Hmm. That’s a sudden adjustment in the reconstruction of TSI dataset?
The steep fall shown in the graphs for the predictions in part VIII is from 11 year smoothing of the PMOD and ACRIM data. And it’s still there in the updated data (see the update in post VIII). But a funny thing happened to the SORCE/TIM dataset. Anthony Watts covered the strange rearrangement of TSI reported on Feb 6 2014. The SORCE / TIM data changed rather a lot overnight. Previously there were four high peaks in the late 20th Century, but now there was only one, and it was the earliest.
Note the dramatic change in the last three peaks. (It’s a 3 second slow blink)
The next data battleground is going to be when the fall in TSI occurred. If it occurred in 2003 as per the PMOD and ACRIM data, then a corresponding fall in temperature is on the cards for about 2017. If it started in 1995 as per the new SORCE/TIM reconstruction now favored by the IPCC suggests, then the corresponding fall in temperature should have been evident from about 2006 — but since it didn’t happen that would mean the solar influence is weak. In the return of a previous theme, the measured data favors the former, while the later relies on reconstructions (the SORCE/TIM data only starts in 2003, and is not relevant to the 11 year smoothed values in the mid 1990s) that flatly disagree with the measured data.
The bottom line here is that perhaps SORCE/TIM is a better guesstimate of TSI than PMOD, but whatever PMOD is measuring seems to be a better predictor of Earth’s temperature. So in future if we can elucidate what makes PMOD useful and the new SORCE/TIM reconstruction not so much, then we may get clues as to the mystery force that operates with an 11 year lag. Obviously temperature is not following the SORCE/TIM reconstruction with an 11 year lag, but it does seem to follow PMOD.
TSI, Satellite Observations – PMOD (At 1-AU distance.)
- KMNI index
- KMNI (PMOD data)
- Download mean TSI by month in W/m2 KMNI: from Dec 1978 to the present (the “raw data” option).
- PMOD absolute scale presumably (the data is from PMOD).
TSI, Monthly Reconstruction – Lean 2000 + Wang Lean Sheeley
- KMNI (WMO FUB TSI Data)
- Download file “, mean TSI by month in W/m2 from Jan 1882 to Dec 2008.
- PMOD absolute scale. Lean (GRL 2000) with Wang Lean Sheeley (ApJ 2005) background. “Spectral reconstruction based on a flux transport model of the open and closed flux using the observed sunspot record as the main input.” At 1-AU distance.
SSN, Observed – SIDC
- KMNI http://climexp.knmi.nl/[email protected]
- KMNI SIDC Data Sunspots
- Download file mean SSN by month from Jan 1749 to the present.
- Or download the same data in a more convenient format at MSFC NASA .
- Data from the Solar Influences Data Analysis Center (SIDC), .
f10.7 Solar Radio Flux at 10.7cm, Observed – Space Weather Canada
- KMNI Solar Flux Radio
- Download file observed radia flux at 10.7cm, by month from Feb 1947 to the present.
- Data from Space Weather Canada.
- Earth-distance (their “observed flux” data; their “adjusted flux” is at 1-AU)
Somewhat post-edited for clarity. – Jo
* Typo: 24th and 25th cycles corrected to 23rd and 24th. Thanks Richard C.
**Typo in 3rd link of Lean and Wang fixed.