Tamino’s radiosonde problem, Part 1

RSS vs. RATPAC tamino

Figure 1. Original found here: https://tamino.wordpress.com/2015/12/11/ted-cruz-just-plain-wrong/

A good month ago, the perennially unsavoury character calling himself Tamino once again tried to hold up the spotty “global” network of radiosondes (weather balloons) as somehow a better gauge of the progression and trend of tropospheric temperature anomalies over the last 37 years than the satellites, by virtue of being essentially – as he would glibly put it – “thermometers in the sky”.

So his simple take on the glaring “drift” between current surface records and the satellites over the last 10-12 years is this: The surface records are right and the satellites are wrong. Why? Because the surface records agree with the radiosondes while the satellites don’t! The radiosondes implicitly – in his world – representing “Troposphere Truth”.

And so, when your starting premise goes like this: the radiosondes = thermometers in the sky = troposphere truth, then any “drift” observed between them and the satellites (as in Fig.1 above) will – by default – be interpreted by you as a problem with the latter.

To repeat Tamino’s fairly simplistic reasoning, then, in the form of some sort of logical-sounding argument: Surface and satellites don’t agree. Radiosondes and satellites don’t agree. But surface and radiosondes do agree. Which means the latter two are right, their agreement robustly verifying the ‘rightness’ of each. (And also, the radiosondes represent “Troposphere Truth”.) Which leaves the satellites out in the cold …

There is, however, a definite issue to be had with this line of argument.

It doesn’t hold up to scrutiny …

From Fig.1 above it is plain to see that there is indeed some kind of “drift” going on. So there is a problem somewhere, no doubt about that …

Tamino at this point performs an interesting test, an idea he got from fellow blogger Eli Rabbett. He detrends the RATPAC (radiosondes) and the UAH/RSS (satellites) series to verify the trivial fact they are all really measuring the same thing: tropospheric temperatures, thereby actually admitting that the satellites, even without being “thermometers in the sky”, still do in fact measure the same thing as the weather balloons do. Tamino’s argument, though, is that, since the “drift” between the radiosondes (“Troposphere Truth”) and the satellites comes on top of this close agreement in actual measurements, a distinct divergence in multi-year progression and trends, then it is necessarily a symptom of some problem with how the satellite datasets are assembled:

tlt detrended

Figure 2. Original found here: https://tamino.wordpress.com/2016/01/15/drift/

Which means that Tamino doesn’t even care to consider whether this problem might in reality lie with the radiosondes, not with the satellites, that the task of assembling a whole bunch of what is basically weather measurement data from individual instruments sent up daily from a handful of separate balloon stations around the world into what is supposed to be a consistent and coherent global climate dataset, might present you with some definite difficulties.

The radiosondes do NOT track the surface!

This is, then, apparently what the satellites are up against, courtesy of Mr. Open Mind himself:

ratpac_raw tamino

Figure 3. Original found here: https://tamino.wordpress.com/2015/09/24/exogenous-redux/

People, I give you “Troposphere Truth” …!

The only problem with the dotted curve in Fig.3 is that it doesn’t agree with anything else. It is clearly wrong.

Let’s have a closer look.

Here’s RATPAC and HadAT2 (a radiosonde network dataset compiled by the Hadley Centre, UKMO) from 1979 to 2015 (the latter one extending only through 2012):

Radios 79-15 a

Figure 4.

Seems a pretty decent fit between the two. But as you’ll notice, the first half of this graph appears rather flat (a radiosonde “Pause”?) while the second half (the satellite “Pause”) is steeply rising. Somewhat peculiar. You see the same thing in Fig.3.

The tropospheric anomalies according to the radiosondes apparently evolved from 1979 to 2015 in an opposite fashion to that of the satellites. During the first 19-22 years of the period in question the temps practically plateaued, while during the last 16-17 years they rose steadily and significantly:

Radios 79-15 b

Figure 5.

We’ll need to direct our focus on each of these two segments separately in order to properly address the problem.

First the 1979-2001 segment, the “radiosonde pause”:

01 02

Figure 6. HadAT2 gl, 700mb, to the left, RATPAC-A gl, 850-300mb, to the right.

Here they are in combination:

1979-2001 radios

Figure 7.

Disregarding that towering 1998 El Niño spike, the trend from 1979 to 2001 all in all looks fairly flat.

Here are the satellites covering the very same period:


Figure 8.

And here’s the surface:


Figure 9.

What’s going on here? Well, we see how tightly the two satellite datasets correlate (Fig.8), and the two surface series (Fig.9) and radiosonde series (Fig.7) also seem to track each other quite closely overall. So what we can do is separately combine each of the two curves in Fig.7, 8 and 9 into a “Radiosonde Mean”, a “Satellite Mean” and a “Surface Mean” to simplify a direct comparison.

Here’s the “Satellite Mean” vs. the “Radiosonde Mean”:

Sat mean vs. radios mean (79-01)

Figure 10.

The “Radiosonde Mean” falls distinctly off relative to the “Satellite Mean” from around 1994.

So yeah, the radiosondes do not agree with the satellites. But we knew that already, didn’t we? That’s the whole point of this discussion, isn’t it?

OK. So what about the “Radiosonde Mean” vs. the “Surface Mean”, then?

Sfc mean vs. radios mean (79-01)

Figure 11.

Well, what do you know! The exact same pattern of divergence as with the satellites! The radiosondes clearly sagging from 1994 onwards.

Let’s confirm this equivalence by comparing the “Surface Mean” with the “Satellite Mean” over this period:

Sfc mean vs. sat mean (79-01)

Figure 12.

Quite a robust correlation, I would say! If anything, the satellites appear to rise ever so slightly relative to the surface from 1996-2001.

If you prefer to have this result spoonfed to you, here are more details:

10 11

12 13

14 22

15 23

16 26

17 25

18 28

19 27

Figure 13. From top left to bottom right: GISS vs. HadAT2 (1979-2001), GISS vs. HadAT2 (12 mo. moving avg.), HadCRUt3 vs. HadAT2, HadCRUt3 vs. HadAT2 (12 mo. avg.), GISS vs. RATPAC-A, GISS vs. RATPAC-A (12 mo. avg.), HadCRUt3 vs. RATPAC-A, HadCRUt3 vs. RATPAC-A (12 mo. avg.), GISS vs. RSS, GISS vs. RSS (12 mo. avg.), HadCRUt3 vs. RSS, HadCRUt3 vs. RSS (12 mo. avg.), GISS vs. UAHv6, GISS vs. UAHv6 (12 mo. avg.), HadCRUt3 vs. UAHv6, HadCRUt3 vs. UAHv6 (12 mo. avg.).

If you want to see the striking difference between the surface (according to GISTEMP LOTI gl mean) and the “Troposphere Truth” (according to HadAT2 gl 700mb) even more clearly, here’s the one superimposed on the other between 1970 and 2001:


Figure 14.

Watch how the troposphere anomaly according to the radiosondes goes deeper and higher when moving from 1970 to 1980, as is (perhaps) to be expected, but how from then on, while the surface anomaly starts climbing steadily towards the 1995-2001 level, the troposphere anomaly according to the radiosondes simply stays put, it completely levels off, so that the two eventually end up converging in the latter half of the 90s. Or, what is actually happening here, as we saw in Fig.10 and 11 above, the radiosonde curve simply goes into a slump around 1994.

It all looks pretty spurious. Like a fundamental problem with how spotty weather data is assembled into a global climate dataset …

A funny aside to this story is how in the late 90s, early 00s, the “Climate Establishment” strongly criticised the UAH team for their low tropospheric trend since the start of their series in 1979; artificially low, it was claimed, an artefact simply of instrument error, or, more precisely, of so-called “orbital decay” of the carrying satellites. The UAH team eventually complied and changed their series to show warming. And so ‘everyone’ was happy, and ‘everyone’ agreed that both the global surface and the global troposphere above had warmed significantly from 1979 to ~2000. Here’s a statement from the US “Climate Change Science Program (CCSP)” from April 2006 about how this “significant discrepancy” between surface and troposphere had allegedly been ‘fixed’:

“Previously reported discrepancies between the amount of warming near the surface and higher in the atmosphere have been used to challenge the reliability of climate models and the reality of human induced global warming. Specifically, surface data showed substantial global-average warming, while early versions of satellite and radiosonde data showed little or no warming above the surface. This significant discrepancy no longer exists because errors in the satellite and radiosonde data have been identified and corrected. New data sets have also been developed that do not show such discrepancies.”

Note how this statement was made in early 2006. And up until some time during 2005, the surface and the satellites (represented by RSS; UAH was still at this time tracking slightly too low) actually agreed conspicuously well; no “drift” to be spotted:


Figure 15. Surface (red) “GISTEMP LOTI gl mean” vs. Lower troposphere (green) “RSSv3.3 gl”, 1979-mid 2005.

However, the radiosonde (HadAT, RATPAC) version of how the tropospheric temperature anomaly evolved from 1979 to 2001 evidently to this day remains way too flat as compared both to the satellites and to the surface (Fig.10, 11 & 13). So when CCSP in 2006 state the following:

“This significant discrepancy no longer exists because errors in the satellite and radiosonde data have been identified and corrected.”

… that is at best a half-truth. They acknowledge that the radiosondes apparently did not represent “Troposphere Truth” back then. The radiosonde datasets had errors in them, and these were identified and – allegedly – corrected.

But were the errors in fact corrected? No, they weren’t. The people constructing the radiosonde datasets rather went to work patching the whole series up to make it seem as if everything was now hunky-dory. As of 2005-2006. And what they did was simply this: They lifted the end segment (the last few years, after 2000) of the radiosonde series to line it up with the end segment (the last few years) of the surface series. Another instance of – recalling Fig.14 – convergence after an extended period of divergence …:


Figure 16.

In other words, they didn’t really correct that part of the radiosonde dataset that created the “significant discrepancy” with the surface records in the first place, the segment which made the radiosonde trend so much lower than the surface trend going up to 2001, and which thus in effect caused the problem in need of solving to begin with …

No, they simply ignored this and opted rather to just lift their series post 2000, so that its final part, leading up to 2005-2006 (inside the light green ellipse to the right in Fig.16), could once again connect with the surface series (as it did until ~1993-95; green ellipse to the left). Always the improvisational approach. Just make the end points agree. As long as the end points link up, what lies in between doesn’t really matter. This appears to be their mode of operation. It’s all about the overall visual effect …

Another and more direct effect – a real bonus one, in fact – of not correcting that part of the radiosonde series that actually needs to be corrected (UAH did, so why not the radiosondes?), that is, by not lifting the obviously slumped 1995-2001 segment, but rather the segment following it (2001-2005) instead, is of course … you guessed it: A much steeper rise in tropospheric temps going from the last half of the 90s to the first half of the 00s!

However, seen from an actual, straight-up scientific perspective, rather than from an ideologically inclined one, what you end up with by always fiddling haphazardly like this with your data, arbitrarily tweaking it to fit – in the present – with some preconceived program, without an overall plan or contextual overview, without any thought given to real physical mechanisms and processes, is yourself painted into a corner.

What you have is a compromised dataset, an inherently inhomogeneous time series that cannot be properly salvaged without a complete makeover, a full reconstruction – a reassembly.

Case in point: Somewhen between 2000 and 2005 you compare the surface record of global temps with the radiosonde data and you realise the latter is tracking way too low. You therefore decide to lift its leading edge – but only that! – in order for the radiosonde time series to reconnect with the surface one. And so, even if the radiosonde 1995-2001 segment is still much too low, making the radiosonde 1979-2001 trend still way too flat, the mean anomaly level of the radiosonde time series has managed to catch up with the surface and as of around 2004-2005 is tracking as high, for the first time since 1993-94. The problem with this approach shoud be self-evident. If someone had pointed out back then – after the lifting of the 2001-> segment – that the real problem here is the spurious slumping down of the radiosonde time series as a whole post 1994-95, and that the entire disagreement with the surface record would be easily resolved by simply lifting this whole segment, then it would’ve made the original adjusters look a bit foolish. They would then have had to adjust the 2001-> segment back down before lifting the whole post 1994-95 segment up equally. So why bother? What they had now, after all, was a rather neat-looking (as in ‘scary’) precipitous rise from 1996 to 2005, much larger than what the proper adjustment would’ve provided them with …

Yeah, so we move on to the second and more recent section of the temperature time series in Fig.4, the 1996-2015 segment, the “radiosonde uphill”:

HadAT2 96-12 RATPAC-A 96-15

Figure 17. HadAT2 gl, 700mb, to the left, RATPAC-A gl, 850-300mb, to the right.


HadAT2, RATPAC-A, 96-15

Figure 18.

There is no “Pause” here. This is the radiosonde “Anti-Pause” … Their pause ended when this one started. Basically around 1999-2000.

So here’s the funny bit, the one that Tamino seems only too eager to ignore, or better, sweep under the rug: From 1979 to 2001, the global surface of the Earth warmed a lot faster than the troposphere above it, as per the radiosonde version of it. From 1999 to 2015, however, the tables are turned; now the radiosonde troposphere all of a sudden warms substantially faster than the surface underneath:

Trends 79-01

Figure 19.

Trends 99-15

Figure 20.

Notice how the GISTEMP trend from 1979 to mid 2001 (Fig.19) is +0.135 °C/decade, while the radiosonde trends are much lower: RATPAC-A, +0.085 °C/decade; HadAT2, +0.071 °C/decade. That’s a troposphere trend just 63 and 53% respectively as steep as the surface trend over the period. Conversely, from 1999 to 2015 (Fig.20) the GISTEMP trend is suddenly much lower than the radiosonde (RATPAC-A) trend: +0.166 °C/decade vs. +0.242 °C/decade (146%).

There is no tenable physical explanation to such a massive reversal in relative warming rates between the surface and the troposphere above it. Rather, this glaring discrepancy is a clear indication of methodological errors

There is much more to this story, specifically related to what happens from ~1995 onwards. But this is for the next post. However, I’ll leave you now with the following two animations for you to ponder; a small teaser, if you will:

GISS vs. H3

Animation 1.

GISS vs. UAHv6

Animation 2.


The radiosonde datasets do not represent “Troposphere Truth”. They rise much less steeply than the surface from 1979 to 2001 and much steeper than the surface from 1999 to 2015. This doesn’t make much physical sense, especially considering how the satellites (UAH) were so roundly criticised for having too low a trend from 1979 to ~2000, how there were such persistent demands of wholesale corrections that were eventually complied with. The errors (the “drift”) pointed out were real, and amended accordingly.

By the satellite teams, that is. NOT by the radiosonde teams …

So the satellites and the surface agree to an impressive extent all the way from 1979 to 2005, the radiosondes constitute the obvious outlier here. The “drift” is between the surface/satellites and the radiosondes, NOT between the surface/radiosondes and the satellites. And so, how will this circumstance affect the discussion of the current “drift”, the one manifesting itself from 2005 ’till today?

More on that soon …

P.S. Someone is bound to argue at some point that what I present in this post not only knocks down Tamino’s presumptions, but also Spencer and Christy’s claim that the radiosondes somehow verify the satellites’ tropospheric trends … And they would be right. I have always found that a strange claim indeed, seeing that it simply isn’t correct. The radiosonde global climate datasets are a mess …

2 comments on “Tamino’s radiosonde problem, Part 1

  1. okulaer: First, what Tamino was responding to was Ted Cruz’s claim about data. As Tamino wrote,”When Ted Cruz said that both satellites and balloon data fail to show warming, he was just plain wrong.. As you’ve shown, the balloon data clearly shows recent warming.

    While there are always issues with any dataset, in this case what’s significant is that radiosondes have not changed significantly in the past 15 years. Averaging them is a relatively simple process and many lay-people have done so. I.e., you can download the original data and reproduce the results. No lay-person anywhere on earth to my knowledge has ever been able to reproduce the satellite data. Richard Lindzen, a favorite go-to scientist for pseudoskeptics said, ““The technical details of satellite measurements are really sort of grotesque.” As a result, almost nothing is simple or consistent with satellite data. Everything has changed over the course of their measurement history. Numerous satellites have failed and even those that worked have had individual channels issues. The atmosphere they’re peering through has also changed. The formulas for calculation have changed numerous times.

    Spencer & Christy are up to what now, Ver.6 Rev 5? And Ver 6 itself has never been published – hence not peer-reviewed. Given their past history of mistakes, are you really going to take anything they do without it being looked at first?

    Meanwhile, Carl Mears, the chief scientist at RSS clearly believes the surface dataset is better. And he felt compelled by Cruz’s circus hearing to write a response. In this he wrote,

    …Senator Cruz concluded that satellite data falsify all climate models, that the planet is not warming, and that humans do not impact climate.”

    This logic is wrong. First, satellites do not provide direct measurements of atmospheric temperature: they are not thermometers in space. The satellite TMT data plotted in Exhibit A were obtained from so-called Microwave Sounding Units (MSUs), which measure the microwave emissions of oxygen molecules from broad atmospheric layers (2-4). Converting this information to estimates of temperature trends has substantial uncertainties. The major uncertainties arise because the satellite TMT record is based on measurements made by over 10 different satellites, most of which experience orbital decay and orbital drift over their lifetimes. These orbital changes affect the measurements of microwave emissions, primarily due to gradual shifts in the time of day at which measurements are made. As the scientific literature clearly documents, the adjustments for such shifts in measurement time are large and involve many subjective decisions. Further adjustments to the raw data are necessary for drifts in the on-board calibration of the microwave measurements, and for the transition between earlier and more sophisticated versions of the MSUs.”

    RSS Version 4 is out. Has been for months. I wonder why you didn’t update with new data?

  2. geran says:

    Kristian, this is WAY off topic so please excuse the intrusion.

    We don’t agree on everything about climate science, but that is not usual for real scientists. I find you about 90-95% correct, as you would probably find me the same. But, Dr. Spencer ‘believes” that CO2 is a heat source. Tony Heller has actually admitted it “CO2 produces warming”.

    If you prove them WRONG, which is easy, you get blocked.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s