BEST is fun

When I said BEST is boring I was primarily thinking of the science. I'm not too surprised to find that many other people aren't. For such folk, there is much fun to be had, so I suppose I'll join in too.

I was going to take the piss out of Watts (h/t KK) for Nature pans BEST and Muller PR antics, prints letter from Dr. Singer, which he wrote in response to a Nature editorial that said

Global warming is really happening -- really. There was no conspiracy or cover-up. Peer review did not fail and the scientists who have spent decades working out the best way to handle and process data turned out to know how to handle and process data after all.

but Watts is dull, so lets take the piss out of Curry instead.

The next bit is really wacky, and apparently evolving as we speak. So its a good idea to start off with some science - Tamino has an analysis of the analysis of the last 10 years of the BEST data. Since that data is essentially the same as everyone else's (as I thought we had now agreed we all knew from the beginning :-) so inevitably it shows the same upward trend (once you remove the obvious broken data; and really, it is obvious, and Tamino even finds the error stats to show it).

Proceeding, again h/t to KK for his A Climate Soap Opera. Which is what it is, so don't read on if you want edification. The Mail on Sunday, which is full of lies, and David Rose, who is full of lies [see end - W], run a story claiming the familiar GW-has-stopped meme can be seen in the BEST data - and they quote the GWPF, who are also full of lies. The GW-has-stopped meme has been around for a while, has been debunked many a time, and the Tamino link above debunks it once again. All very dull, but then Rose gets Curry to say some dumb things, or possibly he just says "wouldn't you say X", and Curry says "yes" (apparently this is the way journos get people to give them the quotes they want). We could do Kremlinology over who really said what, but happily Curry says on her blog "In David Rose's article, the direct quotes attributed to me are correct". Which is nice, so we know that 'There is no scientific basis for saying that warming hasn't stopped,' she said. 'To say that there is detracts from the credibility of the data, which is very unfortunate.' is direct from her. As is As for the graph disseminated to the media, she said: 'This is "hide the decline" stuff. Our data show the pause, just as the other sets of data do. Muller is hiding the decline.. And as you've all read Taminos article, above, we know that Curry is talking drivel.

Why is Curry doing this? Because the only thing worse than being talked about, is not being talked about. And Curry, despite being a BEST team member, was invisible. She is direct about this: "I was contacted by a few journos last week, I made my points, but they were interested in the implications for trend analysis, UHI effect, station quality. I made the point that these were complicated issues, and that I regarded the BEST papers (which were as yet unpublished) to be the first of many analyses on these topics using the new data set. This wasn't what the journos found interesting, and I don't think any of my quotes on this made it into print." And I think she got bored and lonely on the sidelines, and decided she had to say something outrageous in order to get her piece of the action.

[Update: Curry has now resorted to redefining the language in a doomed attempt to rewrite her past nonsense into a favourable light. Tamino in understandably unimpressed]

[2013 update: the link I put in to "" no longer works (I hate it that ScienceBlogs broke a pile of old links, they should know better). The internet archive tells me that the state when I wrote it was: so you can see what I intended.]


* Nick Stokes noticed that the "bad" month only has Antarctic stations in it.
* Quark Soup

More like this

What a shame you censor any attempt to discuss facts. I suppose that is acknowlegdgement you know "catastrophic warming" is a total frauid that cannot survive factual examonation.

I think it was Gavin who once cracked up and bluntly said, "What part of 'short trends are not significant' is too hard to understand?" (or something like that).

Maybe Judy's problem (in addition to being sucked into a conspiracy fringe) is simply a lack of intuitive familiarity with the strange and confusing world of statistics.

Gone emerita.

Sad, really.

By David B. Benson (not verified) on 30 Oct 2011 #permalink

Boring as Watts latest piece may be, its Nature ' prints letter from Dr. Singer' headline is bogus.

Early rising Fred posted the first online comment on last week's Nature editorial, but not even Daedalus or Dreadco could manage the superluminal peer review needed to publish a reply to a leader in the same issue in which it appears.

[Agreed. I talked about that in the original version of my post, but then decided that "Watts lies about something" wasn't notable, and focussed on Curry instead :-) -W]

There's something funky about your second link. It goes to a not found, probably because it ends many characters too late.

[Oops yes, thanks. Fixed -W]

Instead of talking about Judy Judy Judy, which is too easy and goes nowhere, why is nobody talking about Muller's characterization of CRU as being an outlier? Has that just become a given?

[I don't think anyone expects Muller to talk much sense. And everyone is bored by the differences, now, since it is clear there aren't really any of any great interest -W]

What is interesting is seeing Curry being subjected to the same bull***** and harrassment that people such as Phil Jones have been. I hope she enjoys it, acts super professionally, doesn't get annoyed or angry at anyone, realises that it's just normal and expected treatment for scientists, they are, after all, in what is known as a field of endevour noted for it's argy bargy and poor treatment of those who speak out on the science. She should just show good humour, be gracious, not get pissed at anyone. Any deviation from such expected behaviour will be a sign of being secretive, furtive, cowardly, dishonest, conspiratorial, evil, and truly worthy of much more intense attacks, suspicion, and, if I had a say in it, some FOI requests.

"Rose gets Curry to say some dumb things, or possibly he just says "wouldn't you say X", and Curry says "yes" (apparently this is the way journos get people to give them the quotes they want)."

I've experienced this, and it's usually not good reporting. Okay if a reporter is just trying to establish non-controversial background facts, but otherwise the reporter's taking over the story. OTOH, I did actually agree with the reporter and it made it so I didn't have to think of something intelligent to say.

What is notable is that Curry is disagreeing with papers that she is a listed author of, specifically whether BEST shows an arrest in the temperature rise or not. This is a key finding. If this comes to the attention of the editors and reviewers interesting things may happen.

Re #6

CRU is an outlier because, probably an honest error (lets not spawn a whole new family of conspiracy theories here please), for all the other datsets they are showing land averages. But the "CRU" data are HadCRUT which are land and ocean combined. So, of course they are an outlier in the same way a pineapple is distinct from a pink lady, granny smith and golden delicious. Its in no sense a like-for-like comparison and because the heat capacity of the ocean is orders of magnitude greater HadCRUT naturally shows less warming than land only estimates. If that doesn't work think about your toast and your coffee at breakfast.

Anyone who has the time to download the four global land series (all available) and plot even *shudder* in excel can easily verify this. When you change from HadCRUT (pineapple) to CRUTEM3 (pippin) you are suddenly going to be comparing four very similar fruits. Annd CRU will no longer be an outlier.


"But is it? Not according to Prof Judith Curry, a member of Prof Mullerâs team, who claims the same findings have shown that global warming has stopped â plunging the rest of us into a quandary of what and who to believe."

Job done, or is she getting a taste of the Phil Jones treatment? It's anybody's guess.

JB: job done.

It's the certainty of the uncertainty monster. I.E. she's certain that climate science is all hooey, so it's certain there's no warming, because of unknown unknowns which clearly must all fall one way (even though, being unknown, they're unknown, of course!)


that is CRUTEM3V, which has gridbox adjustments for variances, discussed in the CRUTEM literature of I believe late 1990s vintage. HadCRUT3 uses CRUTEM3 (without such adjustments).

The site that you refer to also has BEST analyses that you can overlay. My hunch is that they are not renormalizing to a common climatology so CRUTEM is offset lower but otherwise the two series are in substantial agreement post roughly 1900. For another exercise I am currently plotting the various land series up as global averages and can assure you that the series HadCRUT in the graph Economist et al. have used is not the land dataset. If you renormalize GISS, NCDC, CRUTEM and BEST to a common climatology (1901-2000 say to avoid some sense of artificial agreement over a shorter climatology period) then they basically overplot each other pretty much all the way back to at least the late 19th century.

[Do you have any comment about BEST's decision to start at 1800? From what I've read elsewhere, they don't appear to have extra data to support this. Or can it be defended as a semi-arbitrary decision? -W]

As I stated in the Economist article I think they will need to do a lot of convincing. In much of the early 1800s period the data consists of 30-40 stations in Europe, 2 in New England and 1 or 2 in India. Even if the krigging was yielding a globally complete field estimate I'd have significant doubts about the inferred Southern Hemisphere variations. But its not - the influence is cut off at 1500Km. That also means that much of the apparent field coverage is statistical inference - not real data.

So, the 'global' estimate misses out all our friends in Canada, on the US West coast, in S. America, Australasia, most of Asia etc. etc. I leave whether that makes a viable estimate up to you.

Of course, compounding that is that when you have a really sparse network your probability of detecting any artefacts using a neighbour based approach (which they do) is effectively zero so the data themselves may be substantially biased in this early period.

Probably the most defensible start point choice given current data availability is roughly 1880 when there is sufficient continuous Southern Hemisphere data to both infer SH trends and also have a shot at finding at least blindingly obvious large non-climatic breaks across most of the network.

BUT data rescue and data provision holds out the promise of being able to go back further. You can play around in the surface databank prototype sandpit if you wish where there are pretty pictures aplenty and where you can get an idea of what a significant int'l effort is producing to be released next year. I'd look at the by decade plots in the monthly MAPS directory.

Sandpit here

[Ah yes. and contents? Moyhu has a nice tool at… though that doesn't offer enough time res in the early decades. But it does show the sparsity nicely -W]

Of course if any of your readers have leads to data not in there (one station or thousands) then we'd appreciate hearing from you. Data submission guidelines are available from here. The only way we are going to get a truly global databank is having multiple people contribute.

Apology to WC who is bored with this.

[Don't worry, I'll let you know when I am -W]

Peter -yes, I originally had them together on the graph and it is basically what you describe. I took BEST off the graph before leaving the link as I do not know how to properly compare them with a graph.

Note that WfTs does not have a land only for GisTemp.

What I think they are describing in the news is the BEST group did an apples-to-apples comparison of the three temperature series with BEST to date. At the beginning GisTemp was the problem, but they figured out how to make it an apple for comparison purposes, and at that point GisTemp came out closest to BEST, and Muller describes the CRU apple as an outlier.

Once they finish the oceans and have a global BEST, then they will all be apples?


its just a mistake on their part. They could easily make a like for like comparison by going to ...…

And I would assume the Berkeley group know where their own data is. Its simply a mistake that could be rectified very quickly, and probably will be. The 'outlier' (to the extent that any outlier exists), by the way, is GISS which exhibits less centennial scale warming than the remaining three land products.

Reheated Curried Muller


With regards to the BEST data itself and what it shows. He showed me an interesting graph this is updated from the Rohde article, whereby the BEST data shows good agreement with the GISS data for the recent part of the record. Apparently the original discrepancy was associated with definition of land; this was sorted out and when they compared apples to apples, then the agreement is pretty good. This leaves CRU as an outlier.

Speaking of CRU, Muller related an interesting anecdote about Phil Jones that was apparently related to him by a reporter. When Jones was asked to comment on the BEST papers, he said he no comment until after the papers were published. Maybe Muller was correct in worrying about making sure the IPCC pays attention.

We also discussed problems with the IPCC, Climategate issues, etc., and we tend to mostly agree on all this.

The one disagreement of the evening was over interpreting hurricane data, but that is not something to bother with here.

So all in all, I am ok with what is going on in the BEST project. The PR situation is still a problem, but the media arenât helping here. In any event alot of people are now looking at the data. The BEST team is taking seriously the more serious critiques and are sorting through them. Progress is being made!

Sheesh! Pass me the sick bag, quick.

Peter - thanks for the help. Nobody seems to disagree with you, so I accept that they may have made a mistake.

"Speaking of CRU, Muller related an interesting anecdote about Phil Jones that was apparently related to him by a reporter. When Jones was asked to comment on the BEST papers, he said he no comment until after the papers were published."

Old news, reported in Ghana News Now over a week ago (BBC original). Just an example of Phil Jones practising what he preaches.

"Prof Phil Jones, the CRU scientist who came in for the most personal criticism during âClimategateâ, was cautious about interpreting the Berkeley results because they have not been published in a peer-reviewed journal.

âI look forward to reading the finalised paper once it has been reviewed and published,â he said.

âThese initial findings are very encouraging, and echo our own results and our conclusion that the impact of urban heat islands on the overall global temperature is minimal.â"

A question for Peter Thorne, if he's still around.

I downloaded both the CRUTEM nh+sh/2 and the simple average series. The first gives 0.22C/decade for 1980-2009, and also appears to be the one used in the BEST comparison. The simple average gives a higher trend for that period, 0.26C/dec, only a little lower than BEST and NOAA (both at 0.28C/dec). Intuitively, the simple average makes more sense to me for a "land only" comparison, since it would weight the hemispheres by land area.

Is it possible that BEST did use nh+sh/2 (which is the only one available at CRU), but should have used simple_average?…

(Credit where credit is due - this issue was pointed out by at Doskanale Szare at Moyhu:…

I have a comment about GISS too, but one thing at a time.

Deep Climate,

there was a paper back in 2005, little cited, by Vose et al in GRL looking at what explains the differences between the different surface records. Most relates to the various different ways that spatial averaging (interpolation) and weighting to form a global mean is undertaken. When stripped down to the lowest common denominator approach they were found to be much more similar.

Sadly, there exists no how to guide (if there was we'd be doing something more interesting and this thread would not exist) so there is no definitive right way. Even the choice of spatial averaging can have a large impact and that's ignoring the units of red noise that result from the very real biases that pervade the raw records from a network never designed or managed for climate and which constitute the elephant in the room with regards to trends. If these aren't quasi-random across the network or through time then getting them properly matters even for the global mean (getting them right matters for regional / local regardless).

Of course, with no how to its impossible to in an absolute sense know how well anybody is doing which is where the idea of benchmarking against a consistent set of analogs (answer known a priori) starts to become appealing. At least then you can ascertain how well algorithms work in a 'tame' environment before releasing to the wild hinterlands of the real-world problem-set. That gives you a fighting chance of interpreting the observational ensemble of opportunity at least. It isn't animal farm and not all datasets can be equal.

There's more on the whole end to end process we (look at the site to work out who 'we' are) think we need to do this properly and unambiguosly at and in the BAMS piece linked from there.

Even more fun is the result of Mann's petition to intervene in the UVa v. ATI case. All can be found in the head post at Eli's. Do read the two links I provided there. Most interesting.

But for those who can't be bothered, Schnare (the lawyer for the ATI) is a liar.

By Rattus Norvegicus (not verified) on 02 Nov 2011 #permalink


Thanks for the Vose et al ref. As for your project at, I did look at it quickly (I always like clicking on posters' names). I was aware of this initiative before, but am impressed with the progress made on an ambitious undertaking.

For what it's worth here is my Berkeley-style "land" comparison for 1950-2009. This corresponds more or less to Fig. 7 in the Berkely averaging method paper (although they use 12-month running average). I've added CRUTEM "simple average" as well (dotted line).…

Despite the misleading "HadCRU" label, it does appear that CRUTEM nhsh was used (and so was GISS Ts "meteorological stations"). As expected, though, CRUTEM savg shows better agreement, as the area weighting is closer to the others.

Here is the decadal comparison.

This corresponds to the decadal version of the revised comparison at the Berkeley Earth website.

My emulation matches up very well for BEST, NOAA and CRUTEM. However, GISS Ts has now been replaced with a more appropriate GISS "land mask" version that matches BEST and NOAA, as explained in the notes.