Thursday, May 12, 2016

Jelly Beans: Gavin, Steve, ATTP, Lucia,


What do jelly beans have to do with Rabett Run Eli hears you wondering.  Well, there is currently much ado about jars full of jelly beans, touched off by Gavin Schmidt's deconstruction of the Christy graph.  To say that Christy put his finger on the scale, well, those were pretty heavy fingers and Eli has tried to point to some of them in green, of course

In any case this has touched off a climate audit food fight about something else, both there and at ATTP and at Lucia's and lord knows where else.  The issue has to do with how to compare a single observation to the results of a series of model predictions, FWIW, as ATTP says
In other words, in comparing the models and observations, Douglass et al. assumed that the uncertainty in the model trends was the uncertainty in the mean of those trends, not the uncertainty (or standard deviation) in the trends. This seems obviously wrong – as Gavin says – but Steve McIntrye and Nic Lewis appear to disagree.
Eli had a jar of jelly beans and he showed the jar to the bunnies telling them that they could win a bunch of carrots if they guessed right.  They all guessed how many beans were in the jar, as a matter of fact some of the fatter and greedier bunnies guessed more than once.  There were a lot of bunnies and the distribution of guesses grew smoother as more entries were made.  Given how many bunnies there are, there were a very large number of entries.  Then everybunny gathered about as the beans were counted.

 At the end one of the bunnies won (Thumper), but her guess differed from the well known mean guess by a bit, but the distribution of guesses was very well known.  Another of the bunnies demanded a recount.  So we recounted . . several times.  There were a lot of beans, the table was small, and there were a couple of characters with bulging cheeks, the number counted kept on changing.

17 comments:

andthentheresphysics said...

I posted this on my blog, but I'll post it here too. James has a post that seems to be saying something similar (H/T Dikran). I must admit that I hadn't considered that it was seen as plausible that the mean of the models would somehow represent truth. It seems clear that if the models are reasonable representations of reality, that truth will lie somewhere within the ensemble range, but not necessarily at the centre.

wheelism said...
This comment has been removed by the author.
wheelism said...

Wait - is Eli suggesting that Lucia is NOT Lubos' transgendered nom de plumage?

E. Swanson said...

Gavin brings out more wonders about Christy's work. Gavin's post has graphs from Christy [2016] which claim to include 3 satellite data series for the Global Bulk "temperature". But, there's only the UAH and the RSS versions of the TLT, unless Christy is including the RSS version of Fu et al.'s T24, (aka RSS TTT). There are three versions of the TMT, but is Christy using RSS 3.3 or the latest 4.0, which is warmer than v3.3? Does Christy treat the balloon data to simulate the TMT satellite weighting? As we all know by now, the TMT is contaminated with some input from the lower stratosphere, which is known to be cooling, so the TMT, or simulations thereof, aren't likely to agree with the models at the surface. Then too, what about all those curves from the CMIP-5 runs, are those converted to satellite measurements or are those just the monthly average at the surface? The whole presentation is just another round of the denialist game of Climate Ball and should have been left to be buried in the Congressional paper shuffle...

JohnMashey said...

This graph or variants thereof is popular.

MN Social Cost of Carbon testimony,
PDF p.18 Spencer
PDF p.66 Lindzen

THEN
At 2015 Texas Public Policy Forum, see
Leighton Steward, an old oil guy who is also behind The Right Climate Stuff.

Then Don Easterbrook does so at 33:50 and 34:48.

Hal Doiron, main guy at TRCS, uses one here.

Olof R said...

The satellite record is too short, has unchecked drifts, and the TMT-layer is a mix of troposphere and stratosphere.
If one really wants to know how the warming of the troposphere compares to that of the models, I suggest comparisons like this
I see no reason to discard the models, real world follows the model mean quite well..

EliRabett said...

Yup, What Eli is not sure of is whether Christy is using the TMT or some composite or what ballon series he is using

E. Swanson said...

So, after perusing Christy's testimony again, I noticed that the first graph was actually for the middle troposphere, aka: TMT. His satellite data is from UAH, RSS and NOAA STAR. But which version did he use? The UAH TMT v5.6 has a trend of 0.101 C/dec, while the new TMT V6b5 has a trend of 0.11 thru April. The old RSS TMT v3.3 had a trend of 0.081 while the latest RSS TMT v4.0 has a trend of 0.136. The NOAA STAR TMT v3.0 has a trend of 0.122 thru Dec 2015. Averaging trends from the old version of the UAH TMT with the old version from RSS and the current version from NOAA STAR yields a trend of 0.101 C/dec. Averaging the new newer versions gives a trend of 0.122 C/dec. The point is, either calculation results in a greater trend in the global TMT than the value of 0.091 shown in Christy's Figure 2.

Do we see here another example of Christy's bias against acknowledging AGW?

Windchasers said...

Swanson, I think Christy's chart is older than either RSS v4.0 or UAH v6. But he could do to update the chart.

Something that's rarely mentioned in this conversation is the Look Elsewhere Effect.

Given enough model metrics, you can always slice and dice them in such a way as to find at least one that fails to meet expectations from reality. E.g.,
https://xkcd.com/882/

E. Swanson said...

Windchasers said...

Swanson, I think Christy's chart is older than either RSS v4.0 or UAH v6. But he could do to update the chart.

The UAH v6 series was introduced by Spencer back on 28 April 2015 and they just did an update to v6beta5 in February 2016 which specifically changed the TMT. RSS v4.0 is indeed more recent, having been released on 2 March this year. The main point is, how can one claim to compare data sets without explaining the methods used to produce them? Without knowing his method, his results can't be reproduced. Furthermore, now Christy has expressed a preference for the TMT, even with it's stratospheric contamination, and thus he is disavowing his previous 24 years of claims that the TLT is the superior metric for assessing global warming. I think he needs to explain his change of heart...

EliRabett said...

Whatever it is, it is not TMT

http://www.drroyspencer.com/wp-content/uploads/RSSv4-vs-UAH-MT-original-series.jpg

E. Swanson said...

Eli, yes, the recent TMT v6b5 trend is less than what I quoted. I was looking at the UAH TMT data file after loading it into a spreadsheet to work with the data, but forgot that doing so changed the alignment for the trend columns. The value I quoted (0.11 C/dec) was for the NH, not Global, which, as of April, is up to 0.08. This value is slightly greater than the old TMT v5.6, which is showing a global trend of 0.07. One difference between the two versions is in the Southern Hemisphere, v5.6 shows 0.03 while v6b5 shows 0.06. That the UAH data has trends which are smaller than the other two would explain the trends for the average as shown on the plots...

EliRabett said...

At this point Eli is seriously wondering about the special sauce that Christy is using, if for no other reason that there is no strong rise in the 1998-99 region for the El Nino. The El Nino should have had at least as strong an effect as the two large volcanoes (opposite direction).

Does anybunny have a clue. Eli's best bet at this point is that Christy is somehow faking a point closer to the tropopause

E. Swanson said...

Christy's use of the 5 year average is the reason the 1998 El Nino vanishes. He must have started by calculating yearly averages, then applied a moving average to that new time series. Since an El Nino typically happens around the end of the year, by averaging yearly, he splits the El Nino in half (mol). Next, when he applies the moving average, further suppresses evidence of the bump. One can produce the same effect with a 61 month centered moving average applied to the monthly values, which actually leaves a slight dip at 1998-99. Just another example of S&C's math fun and games...

Bernard J. said...

A lot of people have been very polite about the nature of Christy's shenanigans in constructing temperature trajectories.

Why?

In my corner of the world if an undergraduate (or even a postgrad) stats student submitted a report with these sort of manipulations to put forward a particular notion, they'd be failed. At least, they would have in any of the tertiary institutions in which I've worked - but perhaps their standards are a bit higher than Christy's...

Tom C said...

"2015 was Record Hot'

Yeah, good one Eli. Including that point would have really changed the message.

"Neglects obs uncertainty"

Um, I thought that the observations are thought to have very little uncertainty, which is why they were played up for so many years, until the increase slowed down.

Pathetic

Garima said...

Nice