Another Copenhagen Synthesis Report Error?

Figure 3 of the Synthesis Report of the Copenhagen Congress presents the long-term trend of increasing temperature and the range of IPCC projections.

The sharp-eyed JeanS over at Lucia’s smelt a fish, noticed the trend appeared to be too high, and did a number of replications indicating the window used was 14 years, and not 11 years as claimed in the caption, and also in the list of Figures.

Figure 3: Changes in global average surface air temperature (smoothed over 11 years) relative to 1990. The blue line represents data from Hadley Center (UK Meteorological Office); the red line is GISS (NASA Goddard Institute for Space Studies, USA) data. The broken lines are projections from the IPCC Third Assessment
Report, with the shading indicating the uncertainties around the projections3 (data from 2007 and 2008 added by Rahmstorf, S.).

Figure 3: Changes in global average surface air temperature (smoothed over 11 years) relative to 1990, p. 9

Below is my replication of the figure (using an SSA package in R) with 11 and 14 year periods and the ‘minimum roughness criterion’. The trend lines are displaced by the software, but it makes it easier to see.

rahm11-14

Black are the CRU temperatures, and the linear extensions are the data points appended to the end of the series for the ‘minimum roughness criterion’, which adds a line with the same slope as the slope of last window size (11 or 14) points. You can see that the slope of the added points increases a lot from 11 to 14 point, which explains why an increase in the window size would change the ssa curve.

Blue lines are the ssa, first EOF for the 11 (dashed) and 14 (solid) embedding periods. The removal of the downturning in the 11-window by the 14-window as shown above is clearly shown.

The red lines are the actual ssa forst EOF trends, without the ‘minimum roughness condition’. The MRC is not necessary. This is an arbitrary end condition selected by the authors in Rahmstorf 2007.

So its pretty clear that expanding the window would be necessary to maintain an alarming uptrend in the ssa curve.

JeanS put the question directly to RC.

254. Jean S Says:
Your comment is awaiting moderation.
25 June 2009 at 7:46 AM

Did you change the filter length from M=11 to M=14 in the temperature graph (Figure 3)?

Interesting the no-one at RC has responed to JeanS’s straightforward question yet.

Does the report err in stating in the caption of Fig 3 that the smoothing length is 11?

Did Rahmstorf change the smoothing length in order to maintain the alarming appearance?

67 Comments

  1. Jan Pompe June 26, 2009 1:20 am

    Just the way to make a peak of a sinusoid come a little later :- add some extra lag.

    Very clever but if the cooling continues he’s going to have to add more lag (increase the window period).

  2. Jan Pompe June 25, 2009 8:20 pm

    Just the way to make a peak of a sinusoid come a little later :- add some extra lag.Very clever but if the cooling continues he's going to have to add more lag (increase the window period).

  3. Jean S June 26, 2009 10:27 am

    I re-iterated my question:

    —cut—
    # Jean S Says: Your comment is awaiting moderation.
    26 June 2009 at 5:21 AM

    Seems like my question was somehow lost first time. Here’s it again.

    In #192, Stefan says: “This is shown in detail in our Science paper of 2007, the results of which are shown and updated in the Copenhagen Synthesis Report.”

    So my question is, did you change the the filter length from M=11 (in your Science paper) to M=14 (or something similar) in the temperature graph (Figure 3)?

    http://rankexploits.com/musings/2009/fishy-odors-surrounding-figure-3-from-the-copenhagen-synthesis-report/
    —cut—

  4. Jean S June 26, 2009 5:27 am

    I re-iterated my question:—cut—# Jean S Says: Your comment is awaiting moderation.26 June 2009 at 5:21 AMSeems like my question was somehow lost first time. Here’s it again.In #192, Stefan says: “This is shown in detail in our Science paper of 2007, the results of which are shown and updated in the Copenhagen Synthesis Report.”So my question is, did you change the the filter length from M=11 (in your Science paper) to M=14 (or something similar) in the temperature graph (Figure 3)?http://rankexploits.com/musings/2009/fishy-odor…—cut—

  5. Anonymous June 26, 2009 10:52 am

    JeanS: Should not be a hard question to answer – 11 or 14? From the report:

    This report has been critically reviewed by representatives of the Earth System Science Partnership (ESSP)ii, by the parallel session chairs and co-chairs, and by up to four independent researchers from each IARU university. This extensive review process has been implemented to ensure that the messages contained in the report are solidly and accurately based on the new research produced since the last IPCC Report, and that they faithfully reflect the most recent work of the international climate change research community.

  6. davids99us June 26, 2009 5:52 am

    JeanS: Should not be a hard question to answer – 11 or 14? From the report:This report has been critically reviewed by representatives of the Earth System Science Partnership (ESSP)ii, by the parallel session chairs and co-chairs, and by up to four independent researchers from each IARU university. This extensive review process has been implemented to ensure that the messages contained in the report are solidly and accurately based on the new research produced since the last IPCC Report, and that they faithfully reflect the most recent work of the international climate change research community.

  7. Jean S June 26, 2009 12:57 pm

    I guess I can’t use the comment feature over RC … again comments with newer time stamp are appearing and mine is nowhere to seen ;-)

    • Jan Pompe June 26, 2009 1:10 pm

      Jean S: I think that is a fair indicator that you are right.

  8. Jean S June 26, 2009 7:57 am

    I guess I can't use the comment feature over RC … again comments with newer time stamp are appearing and mine is nowhere to seen ;-)

  9. Jan Pompe June 26, 2009 8:10 am

    Jean S: I think that is a fair indicator that you are right.

  10. cohenite June 29, 2009 5:18 am

    Apparently, Will Steffan has answered Jean S’s question; the filter length was M=15;

    http://www.realclimate.org/index.php/archives/2009/06/a-warning-from-copenhagen/langswitch_lang/wp#comment-127955

    • Anonymous June 29, 2009 5:29 am

      Cool. I will put up a post on the now very relevant E&E article asap. Note the comment here http://www.realclimate.org/index.php/archives/2009/06/a-warning-from-copenhagen/langswitch_lang/wp#comment-128445 questioning Stefan Rahmstorf about why he switched from 11 to 15. Note also it took Stefan 5 days since it was brought to his attention and fess up.

      • Jan Pompe June 29, 2009 6:57 am

        It raised the question though if there are more cool years will he have to lengthen the filter even more to maintain a robust trend.

        I’m not a statistician but I’m wondering if that is a valid thing to do.

        • Anonymous June 29, 2009 7:06 am

          Only if you want to keep the trend going up.

          • Jan Pompe June 29, 2009 7:42 am

            That’s what I though. Thanks

      • Jean S June 29, 2009 8:25 am

        At least in five days time he managed to edit away my link to Lucia’s blog post ;)

  11. cohenite June 29, 2009 12:18 am

    Apparently, Will Steffan has answered Jean S's question; the filter length was M=15;http://www.realclimate.org/index.php/archives/2

  12. davids99us June 29, 2009 12:29 am

    Cool. I will put up a post on the now very relevant E&E article asap. Note the comment here http://www.realclimate.org/index.php/archives/2… questioning Stefan Rahmstorf about why he switched from 11 to 15. Note also it took Stefan 5 days since it was brought to his attention and fess up.

  13. Jan Pompe June 29, 2009 1:57 am

    It raised the question though if there are more cool years will he have to lengthen the filter even more to maintain a robust trend.I'm not a statistician but I'm wondering if that is a valid thing to do.

  14. davids99us June 29, 2009 2:06 am

    Only if you want to keep the trend going up.

  15. Jan Pompe June 29, 2009 2:42 am

    That's what I though. Thanks

  16. Jean S June 29, 2009 3:25 am

    At least in five days time he managed to edit away my link to Lucia's blog post ;)

  17. cohenite June 29, 2009 12:42 pm

    Did I say Will Steffen? I’m seeing that guy behind the bushes; time for a scotch.

  18. cohenite June 29, 2009 7:42 am

    Did I say Will Steffen? I'm seeing that guy behind the bushes; time for a scotch.

  19. lucia July 1, 2009 2:27 pm

    JeanS– Stefan edited out the link to my post where we discuss the fishy odor surrounding his document? Heh. (I only edit out links to porn etc.)

    David S–
    Is the Grinstead extension just a plain straight line? No ‘noise’?

    http://rankexploits.com/musings

    • Jean S July 1, 2009 3:49 pm

      Lucia, yes, it’s just a straight line.

      Here’s how it is exactly done in the code (this is for the “right” end):
      pright=polyfit(idx,x(idx),1);
      This fits (returns a and b) a 1st order polynomial (a straight line)
      y=a*idx+b for M last data points in the series x.
      Then
      paddedX=[polyval(pleft,(-(M-1):0)');x;polyval(pright,n+(1:M)')];
      Appends data x (to right) with values y from the equation
      evaluated at points (n is the last index) n+1,…,n+M.
      Of course, it would be enough to append only M-1 values as the filter length is 2M-1. The last thing can be seen from how the filter is obtained:
      tfilt=conv(E,flipud(E))/M
      (later normalized to sum to unity). The length of E (“maximal eigenvector” from ssa) is M, so the convolution between E and “flipped” (i.e. reversed) E produces a 2M-1 length vector.

  20. lucia July 1, 2009 9:27 am

    JeanS– Stefan edited out the link to my post where we discuss the fishy odor surrounding his document? Heh. (I only edit out links to porn etc.)David S–Is the Grinstead extension just a plain straight line? No 'noise'?

    http://rankexploits.com/musings

  21. Jean S July 1, 2009 10:49 am

    Lucia, yes, it's just a straight line. Here's how it is exactly done in the code (this is for the “right” end):pright=polyfit(idx,x(idx),1);This fits (returns a and b) a 1st order polynomial (a straight line) y=a*idx+b for M last data points in the series x.ThenpaddedX=[polyval(pleft,(-(M-1):0)');x;polyval(pright,n+(1:M)')];Appends data x (to right) with values y from the equationevaluated at points (n is the last index) n+1,…,n+M. Of course, it would be enough to append only M-1 values as the filter length is 2M-1. The last thing can be seen from how the filter is obtained:tfilt=conv(E,flipud(E))/M(later normalized to sum to unity). The length of E (“maximal eigenvector” from ssa) is M, so the convolution between E and “flipped” (i.e. reversed) E produces a 2M-1 length vector.

  22. Jack & Jones September 6, 2009 1:26 pm
  23. Jack & Jones September 6, 2009 1:26 pm
  24. Jack & Jones September 6, 2009 8:26 am