GHCN Adjustments In Iceland–Still No Explanation From NOAA
By Paul Homewood
Before After
GHCN made a tranche of temperature adjustments in Iceland and Greenland in their latest Version 3.1, which was introduced in November 2011. As a result nearly every station in the two countries saw either temperatures in the 1940’s and 1950’s reduced, or temperatures since then increased, or both. The effect of these adjustments is quite stark, as can be seen by the two GISS graphs for Reykjavik shown above. The original temperature record (which has been verified by the Icelandic Met Office) shows that temperatures in the last decade are at a similar level to the previous warm spell around 1940. Under the new version, that warm spell has disappeared with the result that current temperatures appear to be abnormally warm.
On 17th January, GHCN were asked to comment on these adjustments. I have had replies from Bryant Korzeniewski of NOAA telling me that the matter is under investigation, but no further information. The only conclusions I can draw are :-
1) GHCN simply don’t understand how their latest algorithms work and therefore cannot explain how these adjustments have arisen.
2) They are reluctant to admit their mistakes.
(Or both!)
It is worth noting that, according to GHCN’s Technical Report , some of the GHCN software was rewritten last July by a summer student, Daniel Rothenberg and these changes were implemented in Version 3.1. Although, on a global level, the effect of these adjustments is very small, it is clear that the Quality Control procedures at GHCN have not worked properly in this case.
How much longer do GHCN need to explain their adjustments or correct them?
Comments are closed.
NOT A LOT OF PEOPLE KNOW THAT 
The original Reykjavik data which you sent me of course look just like your first plot. I’ve now looked at them in more detail (and Stykkisholmur and Akureyri ) and verified that all these three sets exhibit exactly the same “grand scale” behaviour. Notably the step changes in the early part of the 1920s and in 1965. Only the latter seems to be recognised by climatologists, goodness knows why! I’m wondering whether to contact the Iceland people direct to show them a few plots demonstrating these observations on their data. Can send you some more plots showing just how insightful cusum plots are compared to things like ordinary scatter diagrams or the notably blunt instrument r-squared.!
It will be important to follow up the shennanigans at GHCN, which strike me as being to say the least bad practice, and extremely rude to the data owners, who surely know more about the values than the biased professionals at GHCN. We clearly don’t know what their cheapo student did with his time there, but it looks very suspicious to me.
No need to contact the iceland met office, they are aware of the issue and have already said:
The GHCN “corrections” are grossly in error in the case of Reykjavik but not quite as bad for the other stations. But we will have a better look. We do not accept these “corrections”.
They have started a blog at icelandweather.blog.is which they say will address this controversy.
GISS was using the GHCN v2 raw data. They changed to v3 adjusted data in Dec. 2011. They have an update page explaining the change. I have GHCN v2 up to Sep 2010 due to downloading the Clear Climate Code version of what GISS was using. Here is a graphic to show the huge difference it makes for Rekjavik.
Frankly, I don’t see how GISS can justify using the GHCN adjusted data and then adjusting it again. At least when they were using the GHCN raw data adjustment it seemed reasonable. I suspect numerous stations have changed drastically.
I calculated the annual temperatures Jan-Dec. Any year with a missing month I skipped. GHCN v3 adjusted has more missing data than is found in the raw files.
Paul, I have some GISS graphs from key stations in Scotland which were saved a year or so ago, and didn’t spot any devious data changes when I looked last night. However, it looks to me that Dublin Airport (already a good candidate for UHI when compared with Belfast and Tiree) has been adjusted very recently (early years are not not quite so cold and recent years quite a bit warmer than they were last time I looked). Greenland, Iceland, Ireland… Perhaps someone GISS is working alphabetically through the countries? Sorry, I don’t have the time to look more closely into this.
Lerwick too.
ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v3/products/stnplots/6/65103005000.gif
As Paul says, he and I contacted GHCN four weeks ago about this.
Since then we have received nothing except a few content-free replies saying
“Please stay tuned for further updates.”
Then last week, looking at the GHCN site, I saw that they had changed the code. There is an entry in their CHANGELOG file that says
********************************************************************************
02/07/2012
GHCNM v3.1.1 is released with changes in scripts due to asynchronous execution
problems. Also added ability to add new stations automatically when their
period of record increased enough to process and improved testing when not
enough neighbors to estimate adjustments.
********************************************************************************
So clearly they have found some errors in their code and tried to correct them, but despite the “Please stay tuned for further updates” comment, they didn’t email us about this.
So all the graphs have changed significantly in the last week.
In v3.1.1, for some of the iceland stations, like Stykkisholmur and Akureyri, the adjustments are now not so bad, but for others the new version is worse,
for example in Reykjavik and Vestmannaeyjar. See this picture where a cooling of about 1 degree is turned into a warming of about 1 degree:
ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v3/products/stnplots/6/62004048000.gif
Nick Stokes at his moyhu blog has done a systematic analysis of all the adjustments. He finds that the net effect of all the adjustments is to create a warming of about 0.03C/decade, so that’s 0.3C/century, i.e,. a significant proportion (30% or so) of the claimed warming in the last century is created by the ‘adjustments’.
Just a few minutes ago, I had displayed the GHCN temperature chart for Rochester, Minnesota, among others. With a dozen browser tabs open, I didn’t realise that the chart was still displayed, so I reselected it from the GISS station selector, and displayed it again. Unbelievably, the data had changed by a few tenths of a degree (both for early and late years) within ten minutes! Some kind of official investigation is needed of this unjustifiable shenanigans.
Looks like you’re right. This is the old dataset.
http://data.giss.nasa.gov/work/gistemp/STATIONS//tmp.425726440000.1.1/station.txt
And the new –
http://data.giss.nasa.gov/work/gistemp/STATIONS//tmp.425726440000.12.1/station.txt
2011 has gone from 7.1 to 7.26. Also at the start 1948 + 1949 have disappeared – these were warm years at 7.02 + 7.71. You can see the effect on this GHCN graph.
ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v3/products/stnplots/4/42572644000.gif
The GHCN revision is dated yesterday, so you may just have caught GISS updating today.
Just done a bit more digging. It appears the 2011 change is OK, as the previous version had no figures for Nov + Dec, so was provisional.
The really interesting thing though is the removal of the warm years of 1948 + 1949. I have checked a couple of months with the original station records and these tally with the original GISS dataset. So they can’t justify the removal because the data looks dodgy.
I’ll check the other months and post up.
Thanks
Reblogged this on Climate Ponderings and commented:
Playin with the data
I am also trying to analyse the effects of all this revisionism. The many different versions of the data are rather confusing, so I try to take it step by step at:
http://endisnighnot.blogspot.com/ – see the tab entitled “GISS: Strange Anomalies”
My current thinking is that GISS are depressing older temperatures, inflating later ones, and generating an artificial warming trend. I’m in the process of looking at the fine detail of just one station: Ostrov Dikson, which happens to have an unbroken set of data (no gaps) for August of every year since 1918.
As well as the changes in Dec, major changes occurred this month, both up and down.
Reto Ruedy at GISS confirmed this to me. I’ll forward this over to you.
Paul
Hey! Do you use Twitter? I’d like to follow you if that would be ok. I’m definitely enjoying
your blog and look forward to new posts.
I’m not on Twitter I an afraid.