Bob Tisdale has a new post at WUWT about a supposed “divergence” between temperature at Earth’s surface and in the lower troposphere.
The “money graph” is this one:
Note that the linear trend rate since 2000 for surface data is +0.0181 deg.C/yr, while for lower-troposphere data it’s merely +0.0096 deg.C/yr. On that basis, “the Bob” concludes that the lower troposphere is definitely warming more slowly than the surface, because, as he says, “The average global surface temperature data almost double the warming rate of the average global lower troposphere temperature data during this period.”
But does it really?
What, one might wonder, is the uncertainty in those trend estimates? The data sets show strong autocorrelation, so we can’t just take the naive uncertainty estimates from linear regression. We can approach that issue in two ways: 1) apply an autocorrelation correction, e.g. as in Foster & Rahmstorf, or 2) transform monthly data to annual averages to reduce the autocorrelation to manageable levels. Let’s do both.
Rather than use an average of 3 surface data sets and 2 lower-troposphere data sets, let’s just use one of each. For the surface I’ll use NASA data, for the lower troposphere I’ll use RSS TLT v3. The NASA data give a linear trend rate since 2000 of +0.0188 deg.C/yr, the RSS data a mere +0.0089 deg.C/yr — a “divergence” even greater than for Tisdale’s combined data.
But when we add “error bars” to those estimates, we find that they’re not necessarily different at all!
I’m guessing that either Tisdale doesn’t know how to do it right, or he just saw what he wanted to see and fired from the hip. Or both.
And by the way … when you compute the trend rate since 2000 by ignoring the data before that, even though we actually have that data, you end up fitting a “broken trend” model, yet another foolish mistake.
He does suggest three possible explanations for his supposed “divergence”:
Of course, there are three possible reasons why the global lower troposphere and surface temperature products do not agree with the hypothesis of human-induced global warming:
- First, the global lower troposphere data are flawed, causing warming rates that are too low.
- Second, the surface temperature data are flawed, causing warming rates that are too high.
- Third, the hypothesis of human-induced global warming is flawed, along with the computer models that support it.
How about this: Fourth: Bob Tisdale’s analysis showing this disagreement is flawed. I’ll go with #4.
And while we’re at it, how about #1? The troposphere data sets used by Tisdale are indeed flawed, in a significant way, according to recent revision by none other than the RSS team itself. That’s why they’ve produced a new product, v4, which shows more warming that the older v3. They haven’t yet published an updated “TLT” (lower-troposphere) product, but the best representative of the troposphere is their “TTT” product, which now looks like this from 2000 to the present:
The linear trend estimate it gives is +0.0171 deg.C/yr. Bye-bye to even a hint of Tisdale’s “divergence.”
And of course there’s the much-vaunted balloon data, the one consistently claimed to confirm the UAH TLT data even though it doesn’t. Here’s what the balloon data say, according to the RATPAC (Radiosonde Atmospheric Temperature Product for Assessing Climate) data set:
The linear trend rate, using only the data since 2000, comes in at +0.0281 deg.C/yr, faster than the rate for surface data. But again, some error analysis would be useful before one declares any sweeping jump-the-gun conclusion.
Bob Tisdale ingored the data he didn’t like (the new RSS version 4 and the RATPAC balloon data) in favor of the data that gave him the answer he wanted, he implied a definitive “divergence” when even the data he selected doesn’t confirm that conclusion, and did it so he could imply that “human-induced global warming is flawed.” And, he threw in a snipe about computer models to boot.
That’s my opinion.
This blog is made possible by readers like you; join others by donating at Peaseblossom’s Closet.
Thanks for tackling this so quickly, Tamino! In your copious spare time (!), might you remove ENSO from some satellite data trends, since Tisdale and other fake skeptics seem in desperation to suddenly be blaming ENSO for various things? Ideally, please repeat the Foster & Rahmstorf analysis but on satellite data? That would be awesome, though I have hesitated to ask this since you already spend so much unpaid time on climate research!
[Response: Thanks for the very kind words. Those who want to help support this effort can follow the link (at the end of the post) to donate via Paypal.
Yeah, now that we’re facing the 3rd record-hot year in a row (!) the deniers have resorted to blaming everything on el Nino. I guess I’ll get around to estimating its influence on satellite data … it won’t stop them, but it’ll better prepare us to deal with their denial.]
Back in the days when a denialist favourite was “there has been no global warming since 1998” you couldn’t find hide nor hair of a mention of el Nino in anything that denialists wrote. Mentions of el Nino to denialists were met with stunned silence. I predicted that as soon as a strong el Nino happened again that denialists would suddenly know all about them. They haven’t let me down.
It’s probably important to note the rationale behind the TTT product. If you look at the TLT and TMT channels they includes a significant bleed from the stratosphere. If I understand correctly, the TTT data is merely adding an adjustment to the TMT data to better represent the actual troposphere.
Another point one could make about the selection of the TLT data is, that channel peaks just below the modeled tropospheric hotspot, thus I think it also doesn’t fully represent the troposphere.
“First, the global lower troposphere data are flawed, causing warming rates that are too low.
Second, the surface temperature data are flawed, causing warming rates that are too high.
Third, the hypothesis of human-induced global warming is flawed, along with the computer models that support it.”
I would have modified this as:
1) The data actually match expectations within uncertainty bounds, but, to the extent that LT data is not warming as much faster than surface data as expected, there are a couple options:
a) surface temperature data flaws
b) trop data flaws
c) insufficient time period to pick up signal (e.g., within uncertainty bounds)
d) masking issues (e.g., trop data doesn’t capture poles)
and finally, e) our theoretical understanding of the ratio of trop warming to surface warming is flawed. (which is mostly orthogonal to the question of human-induced global warming).
tamino, hi. While your analysis is correct, as far as this complete amateur can tell, that is, you have an apparent typo in the 2nd para of your post that had me quite puzzled for a while. You said:
Note that the linear trend rate since 2000 for surface data is +0.0181 deg.C/yr, while for lower-troposphere data it’s merely +0.096 deg.C/yr.
But for the lower-troposphere data I believe you forgot to divide by 10 when converting from decades to years. So that should be +0.0096 deg. C/yr rather than +0.096 deg.C/yr. Of course, I might be wrong. And no need to to post my comment in any case.
[Response: Whoops! It’s a typo. Thanks, and it’s corrected now.]
Bob Tisdale “does suggest three possible explanations for his supposed “divergence”:”
And Tamino suggests a fourth.
But both appear to assume that we should expect (if AGW is occurring etc.) that the the land/ocean surface and lower troposphere – two disparate parts of the global atmosphere+land/ocean surface system – should warm at the same rate.
Clearly we should expect some degree of correlation between surface and lower troposphere temperatures, since they are fairly close and exchange energy. But is there an analysis or a model that concludes that we should expect a particular amount of correlation in temperature between these two parts of the Earth? And is that amount 100%, or is it something else?
Slioch, the same thought occurred to me. And of course the work has been done…
There are manytreatments on the Interweb but Science of Doom’s consideration here is worth a look, especially the Manabe and Wetherald graphic at the beginning. Their early model indicated that up to about 12 km altitude there is still warming with increasing atmospheric CO₂ but progressively less with increasing altitude, and above this region progressively more cooling was expected.
It’s a hugely-cited paper, and I’m sure that hours of curiosity can be assuaged by reading some of the refinements to be found in the citing papers list available via the abstract page.
I think Bob Tisdale’s answer #1 should be given greater consideration. Averaging the UAH and RSS series glosses over the potential flaws in the UAH work. There’s evidence that the TLT in it’s earlier forms were seriously impacted by the high elevation ice sheet over portions of the Antarctic. The surface emissions of water are different from that of sea-ice. RSS doesn’t include any measurements poleward of 70S for these reasons, yet, UAH continues to do so. Then too, there’s been concerns that the UAH team mishandled the NOAA 9 merging in the previous version and this adds a warming bias in the earlier portion of the TMT, etc (Po-Chedley’s work). These two differences between UAH and RSS may explain the lower warming trend reported by UAH relative to RSS. I have a report in final review on the subject, which supports Po-Chedley’s work.
There may be other issues with the UAH TLT, but we can’t say much about that, as UAH has not published the official, peer reviewed version of the new TLT/TMT/TP/TLS results. Of course, the entire MSU/AMSU work is based on a theoretical model of the microwave emissions, which I think to be questionable, since there’s the assumption that those emission profiles are the same at all latitudes and seasons. But we know that Mr. Tisdale (and Tony Watts) has no interest in questioning the validity of the UAH analysis.
Thanks Tamino, good work..
However, I think you are using the old Ratpac version that was discontinued after July.2016.
The former beta version is now the official. From Winter 2000 through Summer 2016 the trend is 0.308 C/decade. The latest data point, summer 2016, is 0.973 C.
I hear from time to time about how the radiosonde data compare with surface/troposphere trends, and that results differ with different radiosonde data sets.
J Christy is quoted in this post at Roy Spencer’s blog, referring to his Congressional Testimony, giving the trend from 1979 for RATPAC at less than 0.087C/decade. This result seems to put this data set at odds with model results, but then I see the high trend since 2000 in the article above and wonder if it’s the same data set?
Have you checked out this congressional testimony Tamino? IIRC you’ve posted on Christy in Washington before.