She don’t make me nervous, she don’t talk too much

Obscure, perhaps, but I claim it was by request. My sermon is taken from Removing Diurnal Cycle Contamination in Satellite-Derived Tropospheric Temperatures: Understanding Tropical Tropospheric Trend Discrepancies by S Po-Chedley, T Thorsen and Q Fu, but before I get onto that I need to snark a bit; where would the world be without such?

One satellite data set is underestimating global warming?

I know that one! Its RSS, isn’t it? Wait… you mean it isn’t? Its UAH? I’m confused. And so are the folk in the SS comments. We all know that RSS is the one that “underestimates” “global warming”. You have to be bit careful about exactly what you’re talking about, but lets say its TLT, because that’s the one that is “troposphere” or was close as any of these come to that. Wiki has values up to 2013-13, and HotWhopper has values to 2014… 1.38 UAH compared to 1.26 RSS. So, why are SS picking on UAH?

Well, firstly SS, and Po-Chedley, are talking about TMT not TLT. TMT is “real”, so its better to talk about; but its too high so TLT is more interesting, which is why it was invented, even though its “synthetic”. Refer to the picture, and the wiki article if you don’t understand. But more importantly, they’re only talking about 20N – 20S, i.e. “tropical”, temperatures, which is fair enough – you can talk about any bit you like, but if you do that, and headline your piece “global warming”, and don’t clearly state you’re only talking about the tropics, people are apt to get confused.


Enough dull snark, on with the paper

Yeah yeah. Its about TMT, so lets switch to that. It also points out there’s a third series – Zou and Wang, probably the same as NOAS STAR – which essentially agrees with RSS and disagrees with S+C. Is there anything here other than S+C wrong again?

So, there are various sources of error in a temperature series from satellite, and the most serious of these is believing the idiots who tell you that a satellite based temperature series is simple and more reliable than a surface based series. Cast the bozos aside, and the two most interesting errors we seem to have left when constructing the satellite series, at least in TMT, is diurnal drift and the “NOAA-9 overlap problem”, which turns into the “NOAA-9 warm target factor problem”. Which are both fairly familiar, and both quite well described in the paper, so I won’t worry that I can’t lay my hands on a good description of them elsewhere. The pic helps; you see the little overlap with NOAA-9 clearly enough, which is a problem because the series need to be stuck together with intercalibrations, not absolutely, much like the solar radiometer measurements.

People diss S+C for getting the NOAA-9 calibration wrong, and such people are likely correct, although S+C haven’t admitted it yet. Don’t hold your breath. But its only part of the problem, and this paper is muchly about the other part, E/W drift, i.e. the polar orbiters – which nominally remain at the same solar hour all the time – actually slowly drift with respect to the sun. As the picture shows. Po-Chedley tell me that this drift effect is worse over land, because the diurnal range in skin temperature is higher there (supported by their fig 2, which shows about a factor of 10 difference land/ocean. That same fig also shows minimal corrections over the ocean, which you’d hope would be true, and they take as evidence that their calibration corrections are, errm, mostly correct. Though also that the differences in oceanic correction between the MSU and AMSU indicate that not quite all is well). If you want to correct out the drift changes, which can be larger than the climate changes you’re interested in, you can apply a correction based on a model of diurnal temperature from a climate model (but ZOMG! Models! However, that reaction is stupid, so ignore such bozos) or you can do some weird half-arsed thing (“UAH produces an MSU TMT diurnal correction based on temperature comparisons between three coorbiting satellites carrying AMSU with different local sampling times”) and then “attempt to use [AMSU] during periods when diurnal drift is small”. Weirdly, Christy thinks that S+C’s method is best, and everyone else disagrees.

Simultaneous nadir overpasses (SNO)

Po-C and friends are keen to tell us how they’ve reduced inter satellite calibration problems, by SNO. This seems plausible, but I’m not judging the details: by comparing simultaneous overpasses in polar regions (the (A)MSU, being polar orbiters, meet near the poles every now and again, so presumably can be compared then). So: In this study, we use the most recent version of the level 1C (L1C) MSU/AMSU data produced by NOAA STAR, which was newly released in 2013. The dataset has been described by Zou and Wang (2011, 2013) and The NOAA IMICA calibration is both effective and important to our analysis. There’s a whole pile of other processing description in the paper – do read it, this is J Climate not Nature – one step of which, of course, is to throw out anything too many SD away from what is expected, again outraging the ZOMG-only-use-raw-data Nazis, but they are fools so again, ignore them.

Observationally based technique for removing
diurnal drift biases

So, with – we hope – intersatellite calibration problems minimised we’re left with the drift problem. Because of concerns with the models – I sense, not their concerns, but other people’s concerns; still, its nice to try multiple angles on a problem – Po-C et al. try to develope a different way to tease out the drift corrections.

The actual correction for diurnal drift involves throwing a regression equation at it; I didn’t bother read the details so can’t tell you if its any good or not; most likely the only way to know is for someone who cares to attack it, and see how they respond.

Broadly, from the look of fig 2, their corrections don’t differ strongly from the GCM-based ones. But that broad-brush eyeballing is wrong, as their fig 3 shows.


Their stuff does appear to have done better than the GCM approach, although the difference isn’t huge. See-also their fig 4.

And the winner is

The end result is a TMT trend of 0.115 +/- 0.024 oC / decade, compared to NOAA with 0.105 (though note that this analysis did start out from NOAA’s dataset, before applying their new drift correction, so its by no means independent of it), RSS with 0.089 and S+C with 0.029. So, its clear who the loser is. But remember, this is TMT, not the somewhat more familiar but synthetic TLT, nor is it the surface temperature. They construct a “T24” which is somewhat like TLT in that they try to subtract out the (cooling) stratosphere from the TMT using other channels; and end up with 0.160 (NOAA: 0.149; RSS 0.125; and trailing the field S+C with 0.056). That then allows them to work out the amplification ratio, or troposphere-to-surface warming ratio, which is expected to be greater than 1, and indeed is. Unless you’re S or C.

Spot the deliberate error

Po-C can then plot the difference in TMT timeseries between theirs and others.


Notice the painfully wiggly bits, mostly on S+C. That’s not nice.

Rice terraces in Yunnan

Lovely. Mostly because it looks like a painting, but is actually a picture. My source is wiki commons via Did the Anthropocene Begin in 1950 or 50,000 Years Ago? by David Biello in SA; the original is Jialiang Gao,

While I’m here there’s VV’s Irrigation and paint as reasons for a cooling bias.

I read Safeguarding research integrity in China by Jane Qiu, which wonders why research misconduct is particularly acute in China. Because of lack of rule of law and tolerance of corruption, I’d say.

And that Victor chap is back in Nature. I didn’t bother read it this time; it didn’t seem promising.


* TPP’s take on the same picture.
* Ecology and the environment – ATTP