2011-03-08 20:20:48Possible problem with some reconstruction images here
Ari Jokimäki


A denier was complaining about the images here. I looked at the issue and while the denier made all wrong conclusions about it, I realised that there might be a problem with how the CRUTEM has been plotted there. It seems to have too much warming when I compare it to the CRUTEM data. One thing about that data is that file formats have been quite poorly described. The format is "year, 12 * monthly values, annual value". If I'm quessing right, the images there have been plotted using CRUTEM data for January instead of annual mean.

2011-03-08 21:34:55
Mark Richardson

Maybe... quick eyeball <a href="http://bobtisdale.blogspot.com/2009/01/land-surface-temperature-comparison-by_07.html">here</a> suggests that perhaps the thing isn't right off the scale. Northern Hemisphere only CRUTEM temp increase has been ~1.2 C from peak. And we expect greater warming over the continents, i.e. the land only in Ljunqvist.


Could be a case of January only though. Unfortunately I'm getting ready for a field trip to Sweden on Thursday so I don't have time to do the checks myself.

2011-03-09 03:42:14comment
Robert Way


The Ljungqvist post does not use decadal averages like it should (that's what Ljungqvist uses).

2011-03-09 05:49:13
Ari Jokimäki


I think it might be the same thing with HadCRUT series there as well.

2011-03-09 17:24:38
Ari Jokimäki


Also Figure 1 from here has been criticized that the GISS-data goes too high. Seems that way for me too.

2011-03-17 02:54:49


Sorry to be late responding.  I've not been active on SkS lately, and didn't see this until John kindly alerted me to it via email.

The CRUTEM data shown in the plots that Ari refers to were obtained here:



Decadal means were calculated using the years 1850-1859, 1860-1869, ... 2000-2009.  These were done using all months, not just January (sorry, Ari!).   Those are plotted in the figure centered on the "5" year of the decade (so the mean for 1850-1859 is plotted at 1855). 

The one thing that's questionable in that graph is that, along with the decadal means (which end @ Dec. 2009) I apparently included the "2010s" as the right-most data point, although that was obviously based on less than 1 full year of data.  Looking back, I'm not sure why I did that, or whether it was accidental or deliberate.  Here are the decadal means, in case Ari or someone else wants to check them:

1855 -0.2821
1865 -0.3466
1875 -0.3424
1885 -0.4804
1895 -0.4154
1905 -0.382
1915 -0.3796
1925 -0.1847
1935 0.0025
1945 0.0066
1955 -0.0924
1965 -0.1037
1975 -0.1021
1985 0.1691
1995 0.4797
2005 0.8292

* The 2010 mean was 1.01 at the time I wrote that post.  The final 2010 mean was 0.96.

The only other step was that for each graph the various temperature reconstructions were vertically shifted by whatever constant was needed to put them on a common baseline.  So the specific starting and ending values of the CRUTEM curve differ from graph to graph, but the range is the same.

If these figures are being discussed somewhere in a comment thread on SkS, let me know where and I'll see what I can do to respond to any concerns.

2011-03-18 04:27:25
Ari Jokimäki


Thanks, Ned. It might be that the inclusion of the 2010 value is the problem there, as the difference that I saw there was not that big. The fact that the datafiles there don't have header lines made me check if the January values would produce similar jump and they seemed to do so. This would have been quite natural explanation for that. I'll check the values tomorrow if I get a chance.

It might not be a bad idea to explain this in the figure captions or to make new versions without the last incomplete value. At any case it would also be good to mention that there are decadal means used.