Wednesday, December 24, 2014

Data updates

I've been adding bits, as foreshadowed, to the latest data page, with another rearrangement. I have added the maintained monthly plotter, and I have used it as a framework for updated back trend plots - ie plots of trend from x-axis date to present, that I've used for the recessional of the Pause. It's here, with more detail at the original posts. There is a button to switch between modes - they use the same updated data. There is also a data button so you can see the original numerical data.

I did a calc of the new ERSSTv4 global average for Bob Tisdale's post, and I'll add that to the maintained set.

I have also included the WebGL maps (updated daily) of daily surface data, currently for days of this year, but I may extend. For recent days, it also shows the global average. I see that for the last week or so, these have been exceptionally warm, which balances cold earlier in the month.

So that's probably it for 2014. A Merry Christmas and Happy New Year to all readers.


Posted by Nick Stokes at 8:05 PM

11 comments:

  1. AnonymousDecember 25, 2014 at 6:37 PM

    Hi Nick

    Something is fishy here. Your ERSSTv4 differences look consistent with what is in the paper, but I get something rather different, which I think makes more sense in terms of the known sources of bias.

    Granted, I was asking a different question - what will be the impact on GISTEMP? As a result I added some extra steps: eliminate all ice cells (0.18C), covert to anomalies, difference cells present in v3 and v4, and then calculate a temperature series on the differences. I also included a couple of unnecessary regridding steps for historical reasons. Given all the extra steps I may well have screwed up.

    I need to eliminate the extra regridding steps and then try dropping out the other steps one-at-a-time, to see if I can find out what is going on.

    Kevin C

    ReplyDelete
    Replies
    1. Nick StokesDecember 25, 2014 at 9:05 PM

      Hi Kevin,
      I take it you're referring to what is in the WUWT post (I haven't yet included v4 on my page). What I did there was just to turn -1.8C into NA, subtract 1960-2014 cell means (Bob then changed the anomaly base on the result), and integrate with cos weighting. I checked doing the same with v3b, and got fairly similar result to KNMI, tho theirs was jumpier.

      I've put the code here.

      Delete
    2. AnonymousDecember 26, 2014 at 2:25 AM

      Thanks, that should be equivalent. I'll update my remaining code to work on arbitrary grids and then start comparing steps to dig out the differences.

      Delete
    3. AnonymousDecember 27, 2014 at 8:04 AM

      OK, my results are robust on choice of anomaly period, anomaly method, grid and order of steps. My difference data agree with the graph at WUWT up to 2008. At that point there is a downward discontinuity of 0.06C in the WUWT graph relative to mine - mine shows a nice linear trend from 1998 to the present.

      Next I'm going to try your code and see what I get. At the moment I still don't have any idea of whether the issue is in my code, yours or Bob's.

      Delete
    4. AnonymousDecember 27, 2014 at 9:17 AM

      Got it. The problem is not in the code. The only difference is that I picked up my data a month before you did. With the introduction of the November data, ERSST have introduced a downward shift to all temperatures from Jan 2008 relative to the previous release. I have no idea why, or which is right. Here's a picture:
      http://i.imgur.com/8JP3yA1.png

      There's nothing in the readme. Guess I'd better contact them.

      Delete
    5. Nick StokesDecember 27, 2014 at 3:21 PM

      There is a list of status changes here. The release of 3/10 was beta. No real help with the change that you've noted, though.

      Delete
  2. AnonymousDecember 27, 2014 at 9:59 AM

    And here are difference maps between the old and new versions for 4 selected months:

    Differences for 2007.12
    http://i.imgur.com/xW29r8Y.png

    Differences for 2008.1
    http://i.imgur.com/IL2zdYg.png

    Differences for 2012.1
    http://i.imgur.com/tsy0vuo.png

    Differences for 2014.10
    http://i.imgur.com/CB0ZL1f.png

    Before 2008, they are identical. From 2008 there are a systematic pattern of shifts, mostly positive in the southern ocean and equatorial pacific, negative round a few coastal regions. It looks a little bit like a ship traffic density map.

    The map for the final month is rather different, and looks like a spherical harmonic pattern.

    ReplyDelete
  3. AnonymousDecember 29, 2014 at 11:43 PM

    I just got a reply from NOAA. The version of the post 2008 data were wrong, and have now been updated. The post-2008 data are operational and updated every month, whereas the pre-2008 data are static; an ftp error in testing put the wrong data online. The difference between v3 and v4 now shows no discontinuity. It's still in beta, so fair enough.

    I guess you may want to rerun your code and send an update to Bob.

    Kevin

    ReplyDelete
    Replies
    1. Nick StokesDecember 30, 2014 at 5:56 AM

      Thanks, Kevin. I'll do that. Thanks for following it through.

      Delete
  4. AnonymousDecember 30, 2014 at 7:16 AM

    Hi Kevin and Nick: KNMI just added ERSST.v4 to the Climate Explorer. From 1982 to present it is more in line with the ERSST.v3b data than the Reynolds. I thought something was odd. Thanks for following up with NOAA, Kevin. I'll have a new post at my blog and WUWT tomorrow. And I'll add a note to the old post.

    Cheers

    Bob Tisdale

    ReplyDelete

Subscribe to: Post Comments (Atom)