Monday, April 27, 2015

GWPF Draft

As foreshadowed, I've drafted a submission to the GWPF panel. It's basically a collection of the things I've been writing about in recent months, but structured according to their remit. I was glad to have the incentive to get it all together.

They have required that it be in PDF format, so the draft is here (about 860 Kb). I'd be grateful for any comments.

Update. They suggest less than 10 pages. I'm at 21. They allow multiple submissions, but I think I'll prune the blog posts.


  1. Nick, Kudos for putting this together - though I suspect it will fall on deaf ears (blind eyes?).

    I might remove the final sentence in answer #1 and replace it with something along the lines of: Obviously changes to measurement procedures potentially impact data quality and/or introduce bias. Time-of-Observation, station moves, equipment changes, and the Urban heat Island effect come immediately to mind, but these have been examined, re-examined, and often re-re-examined. "

    I'd be tempted to add something about beating a dead horse, but that probably isn't the politic thing to do :)

    Since I'm unfamiliar with Booker, I'm really not sure what direction they're pointing towards here. Are they really just buying the 'all adjustments are bad' line?

  2. Proofreading:

    Date it. And when you give dates, give the year too. Your plots probably ought to be "Fig 1", etc., and have proper captions. Otherwise its hard to refer to them. The first one doesn't really stand on its own, and it should. It has some incomprehensible labels.

    Its possible this stuff will be printed out. In which case, rather than "link to original post", you're better off with "Original post: URL"(linked).


    "5. Are the adjustment procedures clearly documented... Of course those who like to demagogue these things " I think you've over-used your railing-against-the-terms-of-reference at this point. Either you take them seriously - the panel really are grown-ups, and will think for themselves (in which case you don't need to keep saying this stuff) - or, you're just trolling them, in which case why bother at all?

    In this case, I think the correct answer is to list papers by Menne and Williams, or whoever, and preferrably point them at the bit of AR5 where this is all synthesised.

    "TOBS nailed.... OK, that's a bit triumphalist. Sorry. "

    Again, this is likely to put people off. If you have the patience, I think you'd do better to re-work the postings as least a little so that they flow better. And say "These are reworked to retain substance; links to originals" or somesuch.

    I got bored at that point :-)

    1. Thanks, William. Yes, I think I'll have to make it look more scholarly, with reference lists, figure captions etc. That would include the URLs, and some papers. I'll probably reduce the blog posts to "extracts from". That would save space and I could take out the chatty bits, with the originals available by link.

  3. Actually, having read their "tempdatareview" I have another comment, which is that you'd be better off by focussing on simply demonstrating why the figure they use on that page is so badly misleading. You almost do this - you talk about their figure, but then you veer off slightly into your own blog posts, which no longer focus on the figure.

    Is it worth focussing on that figure? Maybe not. Its not the panels; they could just ignore it. Meh, who knows.

    1. I was ambivalent about dissing the figure - it actually gives me a lead in to talk about TOBS, which I think is quite instructive. And as you say, it's not theirs.

    2. Note that the figure they use is also from USHCN v1 circa the year 2000, back before NCDC used automated pairwise homogenization methods; ironically, the size of CONUS adjustments are actually larger in USHCN v2, so they would have had something more "dramatic" to show if they had done their homework and used the USHCN v2 series that has been out for ~7 years now.

  4. Nick, I've been reading your blogposts. They are not as clearly written as you might think they are, and they don't work as well as you think they do.

    To be a bit specific, you say at times, 'but this has clearly shown to be wrong' and link to one of your blog posts. If a reader follows the link - it is not as clear, or sometimes not at all evident. When I read some of the posts, it required a lot of reading back and forth to come to a point where I can see what your claim was, which means things are not as clear as you assume them to be.

    Secondly, as much as you might despise the fact, it is true that questions have been festering about temperature records and adjustments. BEST did not resolve these issues. Nor did the Muir Russell review, nor NOAA/NCDC. Most of these actors are content crapping on the reputation of the people asking questions and carrying the day. Why do you feel compelled to add to the chorus? Here, I refer to the "oh you seem to be asking questions because Paul Homewood and Christopher Booker - both of whom I consider utter fools - raised some issues'. Guess what - neither are Homewood and Booker fools, nor does your opinion of them matter. Rather, hat happens on the other hand is, by attacking these individuals, you discredit your point.

    Please don't tell me the questions about adjustments are long-settled and people asking questions are indeed fools. I have seen the 'station move' issue with BEST and the whole thing is a complex ball of unstated assumptions. Data-handling is a philosophy, not a science and a done deal. I do not think messing around with station data is a valid step at all, and calling me names will not convince me otherwise. Your submission appears to do so. Cut out all references to people and positions - the submission will shape up much better.


    1. Shub,
      Thanks for the comment. I agree with your point on clarity. I try to write clearly, but when I re-read a lot later, I can see that I often don't succeed. As I said to William, I plan to prune and maybe re-write quite a lot. I have to reduce length anyway, and I can remove some of the more bloggy bits.

      I do get frustrated with something that I also get accused of - wilful obtuseness. In my case, of course, it is steadfast pursuit of truth :). Homewood and Booker have irrational preoccupations, and they make a lot of noise. Those preoccupations show through in the terms of reference, and I was pointing that out. However, I agree with William that there is no point in railing about it here, and most of that will go.

    2. Please don't it to mean I'm bashing on your writing. It was an honest observation, arising from your thread on Eaurn's JC post. I started here, but had to read back and forth for a few jumps.

      On your point #4, you write:

      "This is a question of great apparent interest to Booker and Homewood, but I can't see that it has any
      scientific importance. Regional averages attract little attention, and variations in adjustment effect wash
      out in the global average"

      This type of statement does not strike me as convincing. Many of the datasets advertise and/or provide regional averages. BEST provides single station, regional and continental averages. It actually promotes these products on its website.

      Continent-scale average are pushed in the scientific literature. Steig al 2009 and the accompanying hoopla comes to mind. All the land datasets are essential one giant regional average even though they are global in scope.

      To say Homewood and Booker made up concerns about regional averages does not strike me as true.

      Additionally, the analysis project was set in motion by examination of adjusted local and regional records. I have previously mentioned this - if local and regional averages do not match the regional average, but their adjusted versions do, there is an element of dissonance. If these take place in data-poor regions, the methodology is circular and is to be rejected. The whole controversy is based on discrepancies between adjusted and unadjusted local/regional records.

      Stating blandly that these issues are only of interest to Homewood and Booker leaves me blinking.

    3. If Booker is not a fool, what is he? The only other descriptions that come to mind are far less generous.

  5. Hi Nick,

    I'm hoping to put something together. Since Roman is on the panel it might make sense to give him R code.

    I am sure if I do something it will be over 10 pages, i'd take advantage of the multiple submission option.

    Other guys have commented on the text. It could use a once over.. ha I should talk

    1. Steven,
      I hope you do make a submission - no-one is better placed. For my part, I'm happy to give any code that is wanted, but I wonder if that is how they will want to approach it. I used TempLS to calculate the adjusted/unadjusted difference, and I think I covered most cases that they'll need. He might want to check the code, which is fine, but a fairly large task now. It has a substantial user input file, with lots of options. I could send the code that does the regular monthly runs.

      I think the ten pages is a reasonably soft limit - I think going over would be easier for them than multiple submissions. The text will certainly get a once-over - in fact, pruning for various reasons. I mainly wanted to get a framework, and the feedback has been very helpful.

  6. I've had a chance to skim through your text. It's a really good start. I'd like to read through it carefully and post more comments. I think it is really important to communicate with this panel and not just dismiss them, so thanks for doing this.

    I largely second William Connolley's remarks that you should stick to the facts and take out some of the adjectives and commentary (true as it is!).

    I definitely think you can shorten the "General Remarks". Since the guys on this panel have solid scientific backgrounds, they don't need as much explanation as the broader audience you write to in your blog. In each section of this document, I would just pick one or two of the really punchy figures/points from your current draft and outsource the rest to the citations. Sticking to the primary literature is the way to go. These guys will appreciate the focus and brevity. The key problem with this GWPF review is that it's framed as if nobody has taken a rigorous look into homogenization before. Just listing the detailed publication history of the field can go a long way in service of the science.

    I will try to send in some more specific recommendations if I can manage to find a few free minutes this week. This is my first comment, but I've been reading your blog for a while. Really enjoy it, thanks!

    1. Thanks, Matt,
      Yes, I'll definitely be taking stuff out, and shortening. And I'll make a reference list, with blog posts and some literature. I expect that they will have a list from somewhere, so I won't try to cover everything.