Skip to content

GHCN Adjusting The Adjustments

April 18, 2012

By Paul Homewood




I reported a couple of days ago that not only are GHCN making adjustments to historic temperatures that have been puzzling the national met offices, but that some of these adjustments have been changing almost on a monthly basis. These two towns in Argentina show this up well.


First, Santiago Del. Below are the GISS datasets for 11th April and today, following the April update.






The temperature for 1931 has been reduced by 0.59C , from 19.71 last week to 19.12 this week.  This reduction has been applied up to 1969. (Last week’s version was actually the same as the actual temperature record, i.e. with no GHCN adjustments added.)

The second town is La Quiaca.






Here, the temperature for 1911 has increased from 6.74 to 8.34. Similarly all other temperatures to 1942 are increased by 1.6C (although this only has the effect of restoring the real temperature record, as previously the temperatures had been artificially adjusted down by an even larger amount).

According to Reto Ruedy at GISS

Looking at GHCN’s data for the first station (Santiago Del ..) and comparing it to the unadjusted data, I found that last month that record was not adjusted, this month it was adjusted – that seems to indicate that this station is according to GHCN’s criteria near the boundary of being exceptional and a single additional data point may have caused it to cross that boundary; I would not be too surprised if next month it would switch back to being unadjusted.

Either the original temperature record is right or it is not. If it is right, it should not be tampered with. If it can be proved to be  wrong, and an appropriate adjustment can be justified, the adjustment should be made and then left alone.

Instead the GHCN/GISS temperature analysis is merely the construct of an algorithm, detached from reality.

  1. Paul Matthews permalink
    April 18, 2012 5:19 pm

    Looking back a bit further, the adjustments keep jumping around. For Santiago Del in Jan 1931, you have April 25.7, March 26.3.
    In Jan it was 25.6 and last Nov it was 25.9.

    For Quiaca Jan 1931 you have April 13.8, March 12.2.
    In Jan and Nov it was 13.1.

    The other weird thing you can see from your tables is how the ‘gaps’ in the data move around. For Santiago Del there was a gap in the data in 1956, but now the gap is in 1972! It seems that another ‘feature’ of the GHCN adjustment algorithm is that when it puts in a step change to the adjustment it deletes some data around that point.
    When most people look at the GISS graphs and see a gap, they would assume that means the data is missing. In fact it usually means GHCN have put in an adjustment there.

    Reto Ruedy’s explanation does not make sense – adding one data point in 2012 causing the whole record from 1931 – 1969 to change by 0.6? There is nothing exceptional about this station, as we have seen that similar things happen elsewhere.

    The only explanation is that there are serious bugs in the GHCN adjustment code, which is what you and I have been saying for the last three months now.

    • April 18, 2012 6:27 pm

      Regarding the “gaps”, at La Quiaca there is a full temperature record in the 1940’s in the “actual dataset”- not one single month missing. And yet in the “adjusted version” the whole of 1943-45 is “missing”.

      This graph is the “unadjusted version” ). There is nothing abnormal happening around 1942. If any adjustment was necessary, I would look at the missing data from 1994-2003. After this break, temperatures shoot up. Regardless of what the algorithm says, breaks like these should be automatically dropped from the dataset, as they introduce the possibility of corrupted data.

      I get the impression that Reto and co are so immersed in their algorithms that they have forgotten the real world.

  2. Brian H permalink
    April 19, 2012 7:37 am

    Nothing like mysterious undocumented algorithms to manage the climate! Where would GISS be without them?

Comments are closed.

%d bloggers like this: