Skip to content

More On GISS Tampering

January 16, 2013

By Paul Homewood


Fig A2


I mentioned yesterday that I had noticed that GISS had, in the latest monthly update, adjusted up previously declared 2012 monthly temperatures by up to 0.06C. September, for instance, had changed from 0.60C to 0.66C.

It was a rushed post, as it was hot off the press, (and it was also dinner time!) I am now able to fill in some detail, and run some comparisons with original data going back to 1998.


First though, I mentioned that I had contacted GISS to get their comments. Reto Ruedy there got straight back to me with this reply. (For all my moans about GISS, I must complement Reto, who has always responded quickly to my requests in the past, which is more than I can say about NCDC!)


Hi Paul,

Last week Dr. Hansen decided to switch to ERSST for the ocean temperatures rather than using our Hadley/Reynolds combination. We will write a note concerning that change with an analysis of its effect.

However, since the margin of error for these estimates even for the most recent years is about +-0.05 C, differences of the order 0.03 C are well within that margin of error, hence not significant.

I personally like the switch, since now all our basic data (land and ocean) come from the same source which simplifies the description of what we are doing.

Thank you for your interest,

Reto Ruedy


I think this raises the following points:-

  • It confirms I was right!
  • Given their previous margin of error was 0.05C, I guess we can now say it is 0.08C
  • On a more serious note, they have not given any reason for abandoning Hadley/Reynolds, or suggested there are errors or inaccuracies with it. This leads to the inevitable conclusion that the Reynolds dataset did not give them enough warming, so they looked for another set that did. If the data does not fit the conclusions, change the data!
  • I also made the point that GISS had not issued any updates to explain any changes. I note that Reto says this will now be done, but I don’t believe it is acceptable for this to be done after the event.

Effects on 2012 Temperatures


The table below shows the originally declared temperatures for last year, with the now revised figures.


Month Original Revised Diff
Jan 0.32 0.36 0.04
Feb 0.37 0.39 0.02
March 0.45 0.49 0.04
Apr 0.54 0.60 0.06
May 0.67 0.70 0.03
Jun 0.56 0.59 0.03
Jul 0.46 0.51 0.05
Aug 0.58 0.57 -0.01
Sep 0.60 0.66 0.06
Oct 0.68 0.70 0.02
Nov 0.69 0.68 -0.01


The average effect for the 11 months is 0.03C. As I mentioned yesterday, similar changes appear to have been made in earlier years, as far back as 1998. Bear in mind, though, that these are anomalies against a baseline of 1951-80, so the baseline years cannot have been affected in relative terms. I can now fill the gaps in some of these earlier years.


Modifications to Temperatures 1998-2011

I have managed to track down the original data, as published by GISS, for most of these years. I have also obtained the GISS data, as they were showing it in 2008. This set was archived by Climate4You. The dataset is available here.

Putting these two sources of information together, we can see how far GISS have modified the data since.


Met Year (Dec-Nov) Original 2008 Version Change Original v 2008 Version Latest Version Total Change Original v Latest
1998 0.60 0.57 -0.03 0.61 0.01
1999 n/a 0.34   0.40 0.06
2000 n/a 0.34   0.41 0.07
2001 0.45 0.45   0.50 0.05
2002 0.51 0.57 0.06 0.61 0.10
2003 0.51 0.52 0.01 0.57 0.06
2004 0.48 0.50 0.02 0.54 0.06
2005 0.60 0.62 0.02 0.63 0.03
2006 n/a 0.53   0.58 0.05
2007 0.57 0.60 0.03 0.65 0.08
2008 0.44 n/a   0.48 0.04
2009 0.57     0.58 0.01
2010 0.63     0.67 0.04
2011 0.51     0.54 0.03


So it appears there have been two tranches of modifications.

  1. The latest switch to ERSST, which seems to have added about 0.03C to the original temperatures.
  2. Earlier changes that added between 0.01C and 0.06C to temperatures between 2002 and 2007. (Whatever these changes were, it is reasonable to suppose that they also apply to temperatures since 2008).


So, what conclusions can we draw?

  1. It is noteworthy that 1998 had been reduced by 0.03C by the 2008 version. This, of course, affects comparison of 1998 with years such as 2010 and 2005.
  2. How significant is a change of, say, 0.10C? Well, compared to 1980, latest GISS temperatures show an increase of 0.24C, so nearly half of this is represented by “modifications”. And this is before we look at margin of error!


The 2008 version, that Climate4You archived, has the full GISS data back to 1880, and suggests that earlier decades have subsequently been cooled, thereby doubling up on the warming effect of adjustments to recent years. But more on that tomorrow.



Original GISS reported temperature available below.

  1. Ray permalink
    January 17, 2013 10:22 am

    The individual adjustments may be within the margin of error, and therefore not significant, but when you add up all of the changes, it’s funny how they always increase the apparent warming trend.
    Hadley/CRU also reduced the anomaly for 1998, via their switch from HadCRUT3 to HadCRUT4.
    Whether this is more accurate or not, it does suit their purposes that 1998 was no longer the warmest year on record and 2010 was.
    I am not given to “conspiracy theories”, but this does make me suspicious.

  2. Andy DC permalink
    January 18, 2013 12:45 am

    The ratio of upward adjustments to downward adjustments speaks volumes for itself. In the world of real science, that type of bias would simply not exist.

  3. Brian H permalink
    January 31, 2013 1:16 am

    Pay no attention to that man behind the curtain.

Comments are closed.

%d bloggers like this: