Pages

Friday, May 22, 2015

Wood for Trees: You Can't Trust It


Added 6/8/15: Wood for Trees has been updated, per a comment below. The developer and maintainer asks that, if you like WFT, you consider leaving a tip here:
https://www.justgiving.com/WFT/

A lot of people like linking to graphs on Wood for Trees, for plots like the one at the right.

Except WFT is using some bad data.

Here are the raw data used to make the chart on the right.

The problem is, these data only go up to decimal date 2014.5, when the current HadCRUT4 data goes up to March 2015.

Second of all, their anomalies aren't current. Here are their first three data points for the graph at right, starting in 2010:

2010 0.545
2010.08 0.558
2010.17 0.668

Here are the actual current first three data points they should be using:

2010.04 0.559
2010.13 0.572
2010.21 0.677

The WFT raw data output says their data comes from http://www.metoffice.gov.uk/hadobs/hadcrut4/data/download.html . Problem is, right now that page returns "Not Found." 

So where is WTF getting their data? And why isn't it current?

Then there's the small problem of their decimal dates. For January 2010, for example, they should be using the mid-point of the month, instead of the beginning. 

So in the list up above, "2010" should actually be, for January (month=1) of that year, 2010 + (1-0.5)/12 = 2010.042. It's a small error, but it's still an error -- it will lead to end-point errors.

--

I have never really trusted Wood for Trees, and never cite it. (I prefer to download the data and calculate for myself.) For example, their Credits page doesn't specify which version of UAH they are using, though its seems at most to be v5.6, when UAH is now on v6.0beta2. Perhaps they are waiting to see if that version persists, but not long ago I noticed WFT offering UAH v5.5 when v5.6 had been out for awhile.

Wood fore Trees is a good idea. But it doesn't look to be maintained or current. And, as the above shows, it's not even clear where it's getting its data from. 

If someone quotes you something from WFT, be skeptical, and double check what data it's using.

14 comments:

  1. I always use the links from www.climate4you.com

    ReplyDelete
  2. Climate4You is an openly denialist website founded and run by Ole Humlum. It has zero data credibility.

    ReplyDelete
  3. I was only saying that I use the links from www.climate4you.com to get to the data from NASA GISS, NOAA NCDC and HadCRUT4. I made no comment on the veracity of www.climate4you.com itself, only that it is a good source of links to the data.

    ReplyDelete
  4. Ned W7:22 AM

    Mallett was posting a lot of nonsense over at Tamino's, until a week ago when Tamino showed him the door.

    Not surprised to see him promoting junk-science sites like Humlum's.

    ReplyDelete
  5. I was promoting the links from www.climate4you.com (it's a very useful 'hub' to use to find the various data sources) - not the content of the site itself. I always use the data from the sources to plot my own graphs.

    ReplyDelete
  6. well it looks OK now, maybe they updated. The link you give http://woodfortrees.org/data/hadcrut4gl/from:2010/plot/hadcrut4gl/last:60/trend goes through 2015.25 and has the same anomalies you say should be there.

    ReplyDelete
  7. Apparently Richard is a Mathematician. That's a revelation.

    ReplyDelete
  8. David Appell

    As far as I know, the owner of WfT is acting in good faith, unpaid and largely unthanked for his efforts. In the past, when I was making heavier use of the site I emailed the owner and asked - politely - if he would be kind enough to update some obsolete data links (same situation as at present). He did so, and I made a donation to the Woodland Trust. Might I suggest that this is the ideal MO going forward?

    ReplyDelete
  9. J Bowers5:28 PM

    Wood For Trees has been updated. Time for a Wood For Trees Can Be Trusted post, yeah, with an edit link at the top of this post and perhaps a link to how the one, single bloke maintaining the site can be financially compensated? A bit like Comical Tony gets compensated via his Paypal button, Heartland Institute $70,000+ website creation that could be done for low four figures, and weather instrument sales. Just sayin'.

    ReplyDelete
  10. Anonymous8:20 AM

    Yes, very disappointing this is not updated.

    ReplyDelete
  11. Thanks for the update -- I've noted it at the top of the page.

    My complaint wasn't with the site's developer/maintainer, though I can see how it came across as that. (My posts and comments often sound more harsh than I mean them; it's a problem of mine that I'm working on.) It was more with users who toss WFT graphs around without looking into the details of the underlying data -- like version numbers, updates, etc. I always prefer downloading the data and doing my own calcuations and graphs -- then I know exactly what data I'm using and exactly what I'm calculating.

    ReplyDelete
  12. I tried it yesterday, but I couldn't get it to display the graph and the trend at the same time. In any case, it doesn't (AFAIK) give a value for the trend, so I always get the data from the links at www.climate4you. com and plot it in Excel, which is very easy. Probably the free Office programs like Open Office and Libre Office would do much the same.

    ReplyDelete
    Replies
    1. You need to open two instances of the time-series you are examining, one for the curve and the second for the trend. Like this. You can get the annual trend from the [raw data] link below the graph display. Scroll right to the bottom of the numerical data and look for something like this:

      #Least squares trend line; slope = 0.00678428 per year

      WfT is much quicker than downloading the data into Excel and has the great merit that the results are both absolutely transparent and instantly shareable online.

      Delete
  13. David: I know this is a long time coming but I have now updated WFT to use HADCRUT4.4, which both brings it up to date and should fix the data changes you noticed. The problem arose because when Hadley update their version they change the filename, and the old dataset (4.3) stops, so one of the few things I have to do manually is update the FTP script, and you know, life happened... :-)

    I understand the point about the decimal dates, maybe mid-month would be more technically accurate, but can you explain what errors you think that might create?

    Best wishes

    Paul Clark
    WoodForTrees author

    ReplyDelete