Monday, November 7, 2011

GMST trends - a cherrypicker's guide.


12 comments:

  1. I think the proper technical term to describe this kind of post is "awesome". :)

    ReplyDelete
  2. Yes, it looks like the sort of thing which would be very informative for the non-chromatically challenged :P
    Kevin C

    p.s. Only joking. Don't waste time doing a mouseover version for me.

    ReplyDelete
  3. I agree with Toto and Kevin C, very nice job.

    ReplyDelete
  4. Thanks, guys. Kevin, that's actually a very good idea. To make the left legend the active one, so a particular level stands out. Helpful for everyone, not just with color issues. I could only do one index, though.

    ReplyDelete
  5. I did something similar back in September, only using start year and length-of-trend instead of start and end year. I think our methods are almost identical, but the results get tilted by 45 degrees. These plots are indeed a great way of picking out cherry-picks. I'm not sure which plot is more intuitive.. I think your way of plotting is more useful for picking out dates that have an outsized influence on trends, but mine is a bit more useful in seeing that trends average out to warming over longer time periods.

    Example

    Cheers, and nice work,
    ----Paul from VA

    ReplyDelete
  6. Paul,
    Yes, that's well done. For many purposes your combination of variables is best. There are three things of interest - start, end and duration, and you can put any two of them on the axes and the third ends up as 45 deg lines. In my previous post, which got me started on this, I had end date on the x-axis and duration on the y-axis.

    I thought here that end date would be on people's minds, so should go on the x-axis. Duration on the y-axis has the advantage that for the shorter time ranges, you don't need the triangle cutoff and can use the whole space. But I needed space for the legends anyway ;)

    ReplyDelete
  7. I just downloaded the "BEST all land monthly averages" with the link on this page. I have searched for these BEST temperatures for some time, and I am happy to have found them on your blog. On the Berkeley web page I have only find the temperatures for all the stations; much data that need to be processed before giving any meaning. Have you received your temperatures from Berkeley, or have you processed them yourself based on the stations temperatures. If you have processed them yourself, why does not Berkely make the temperatures available in the same human-friendly way as you have done ?

    ReplyDelete
  8. Hans,
    Yes, the BEST web site is not well set out for finding things. I used just the monthly averages compiled by BEST - the file called Full_Database_Average_complete.txt in analysis-data.zip.

    I have downloaded the big data set, and I'm planning to put it through TempLS, which is my least squares program. I've been holding off because the data has been de-seasonalised in some unexplained way. It shouldn't affect my analysis, but I'd like to know. Also it's about the time of month when I usually download the new GHCN and SST data to produce the October mean and plots, tho this month I may be late.

    Robert Rohde of BEST was noted for his work with GlobalWarmingArt, so I'm expecting there will be better graphics from the team when he gets time.

    ReplyDelete
  9. Nick,

    The de-seasonalising in the current Preliminary data seems to be work in progress.
    The current released analysis code simply discards data that seems not to be properly processed (too high std deviation):
    (from BerkelyAverageCore.m)

    %%%%%%%%
    % Temporary Fix: Remove ultra-high variability created by bad seasonality
    % procedure on sparse data. This will be removed in the future versions
    % when the seasonality procedure is fixed.
    parfor k = 1:length(se)
    data = se(k).data;
    if std(data) > 7
    orig_map(k) = false;
    end
    end
    %%%%%%%%

    ReplyDelete
  10. A year gone by and Steve still has not redone his messed up calculations on MMH. Or opened the thread where he locked comments after finding he messed up.

    ReplyDelete
  11. I'm a little disappointed I've missed this until now. Great work Nick!

    ReplyDelete