Stykkisholmur (Iceland) temperatures: From reality to GHCN v1

Guest post by ducdorleans who is a Belgian civil engineer.


The Iceland Met Office (IMO) maintains some of the most carefully curated climate records in the World. This does not stop the Global Historical Climatology Network (GHCN) from tampering with the data, normally in a way that cools the past making today appear to be anomalously warm.

The winter of 1929 was an unusual one in northern Europe. In February 1929, for example, England was gripped by frost while Iceland was uncommonly mild. It was in fact 17˚ warmer in Reykjavik than in London. GHCN overcame this inconvenient historic warmth by simply changing +5.4˚C to -5.4˚C creating an 11˚C swing between reality and the version of history preferred by climate science. A forensic analysis of the weather station at Stykkisholmur is provided below the fold.

1. GHCN v1

This post was inspired by and builds upon Re-writing The Climate History of Iceland published on Energy Matters in February this year.

Bringing together, into GHCN v1, all (or most) available information from the hundreds of National (and maybe some private) Meteorological organizations from around the world – as done by The Argonne National Laboratory around 1990 – was a most commendable task. I can only begin to imagine the then workload.

Some individual MET’s had surely started their own GHCN-like task of bringing together and homogenizing their own meteorological data for temperature, pressure, precipitation and others. With some explanation on how their Fortran or Cobol generated data files were structured, and maybe some data quality control, these data were ready to be put into the Big Global GHCN database in a minute.

But others might just have come into the form of a printed book, or a handwritten ledger of records, with 1978 just after 1961, and just before 1945. Hard to recognize handwriting maybe? Diminishing ink?

This is what GHCN themselves have to say about it in the accompanying document for the GHCN v1 data.

All in all, not only commendable, but surely a Gargantuan task ! But also an extremely important one ! GHCN v1 is the original database, on which every other future version would (or should, but as it turned out “might”) base itself on.

2. Icelandic meteorological data

I got interested in the Icelandic temperature data because I read, on Paul Homewood’s blog, that the “original IMO data were available online”.
“Original data? … online!? “ … WOW!

I remembered – reading e.g. here – Dr. P. Jones’ difficulties to get by “original data”. The struggle his department had with the national METs about them. The treaties between Governments and States over their use. And the difficulties to properly archive them, so that, if somebody asked for them a few years later, they were not lost in an office move.

And now, for Iceland, they would simply be available online ? …
A real climate science luxury!

3. Icelandic Met Office (IMO) website ( )

In view of the ongoing controversy over past temperatures, the IMO website is, as to historic data, what a meteorological website – IMHO – should be. Clear, open and transparent. Even the written or printed published records since 1873 can all be searched online (publications section of the about page)!

You are interested, for some strange – or just some – reason, in the temperature observation at 9pm on 21st October 1873 in Stykkisholmur ? … Well, it was -1,4°C, the highest observation of the day!

If you want to know some monthly values, no problem either … Interested in the observed average temperature during the month of March 1929 in Stykkisholmur? … This page is giving the answer: it was 5,4°C (it is on “Blaðsíða” 10 …)

4. Longer IMO temperature series

Apart from a model website, (at least some of) IMO staff don’t need FOIA rules to offer more help and information either. After my first looking into the different data sets, and with a faulty conclusion, I was wondering about them and sent an email to Trausti Jónsson, meteorologist at the Iceland Met Office. A prompt reply, and links to more information followed immediately. Iceland and Icelanders are taking their climate very seriously. No feet dragging, no formal request I had to put in … complete transparency, as it should be.

His reply email explained some – e.g. :

“Most of the observations made before 1870 are generally considered to be of a lesser quality than those made later – and some of them are. The dmi published icelandic data in their yearbook. We have made those accessible online. However, most of the temperature averages published there are not entirely compatible with later averages because of changes in observing routines – this is the main cause for the adjustments that we have made on the series.

The Stykkisholmur series proper starts in November 1845. The homogenisation of this was done in 1970 and the results published in 1975. There are no later adjustments to the Stykkisholmur series.”

(As a side note, what I never understood, both as a civil engineer and a citizen is that any data, in possibly the most lethal and most dangerous challenge mankind has faced in modern times – i.e. CAGW (catastrophic anthropogenic global warming) – , could possibly be hidden behind a paywall or FOIA rules. These data – especially these data ! – should all be publicly available at no cost and at all times. Because of all the possible dangers involved, both in human lives and economic losses, withholding any of them, or offering them at a price, should be a criminal offense !)

5. Why original data?

It is obvious that all and everything should and has to be compared to the original observations.

All over the world, there were the men and the woman who went out in the rain, heat, cold, the night, the day, to read that thermometer or that pluviometer, and what would possibly be their reason to NOT do it properly. I’m absolutely convinced that they took great pride in being the ones to have to do this work.

But, those original observations are not necessary the gospel. Maybe, the procedure – what they were told to do – could have been, looking backwards, sub-optimal and really should be adapted. And therefore the original observations might have to be amended, and/or corrected.

In Iceland, that was all done by the IMO starting 1970, as Mr. Jonssón wrote … “After that, I suppose that no other “homogenizations” would be necessary …” On his blog, he writes:

“Everyone can look up the numbers and compare both adjusted and earlier published values. Most of the later adjustments that have been made by the IMO are concerned with changes in the observing hours that necessitate slight adjustments. The methodology of the current calculation of monthly means is described in detail in:

Exactly the same method has been in use since 1956. There is no mystery surrounding the data from the Icelandic Meteorological Office. The current version (2003) of the Reykjavík data is available at the website of the IMO (and has been for many years):

and the Stykkishólmur series has not been adjusted since 1975 – only updated:”

6. About the data, the used datasets, and the stations …

a) Data


World Weather Records, “published since 1927“


leads to “” which I used – 23,103,900 bytes (accessed 11/03/2015)

GHCN v2 Temperature files in the GHCN v2 database seem to have gone walkabout atm …

( = “raw GHCN data + USHCN corrections”)

GISS v3 hom.
IMO long
IMO monthly

(Trausti Jónsson’s blog … this is a tabled summary of )

  • In july 2012, Mr. Jónsson posts “Temperatures in Iceland – as originally published” ( ) in which he gives other temperature data in the file “original_pub_temperature_values.txt” , linked with the stations whose metadata are in the file ” original_pub_temperature_stat_info.txt ” … these temperature data, I called “IMO unaudited” (imo U)
  • So, I have 3 datasets – GHCN v1, and IMO “audited” and “unaudited” – with sometimes overlapping time frames.

b) Stations

( * = NO “audited” temperature data )

7. IMO vs GHCN v1

Ok, then … Since Stykkisholmur temperature data are available, in the GHCN and IMO datasets, for minimum the years 1846 to 1980, let’s compare them …

For all the months of the year, except January and March, hardly any differences … Both datasets seem to have been recording the same variable !

But what about January and March ? …

Up to nearly 12°C difference in certain years ? If we add those 2 months together, this is the result:

From beginning of the 1920ies up to the end of the 1940ies, GHCN v1 combined January + March monthly values are up to 15° colder than the IMO data. What is at work here ? Some highly scientific, wrought and otherwise profound algorithm that presumes GHCN v1 values more worthy then IMO values ?

Not really, just a change of sign ! The above mentioned REAL +5.4°C in March 1929 in the IMO dataset simply becomes -5.4°C at GHCN v1. (a quick sidestep to today: To make matters worse, in the GHCN v3 dataset, the real average temperature in Stykkisholmur has been tortured more to now become an even colder -5,71°C !)

Check it out in the original data, or in the accompanying excel sheet.

The case of March 1929

a) March 1929 in the original yearbooks

b) January to March 1929 temperatures in Iceland in the newspapers

9. Discussion

Everybody can make mistakes in manually inputting values into a database. However, the methodical nature – only in January and March, and only in relatively warm months – of the error makes it less and less of a mistake, and more and more of an effort to slightly change the yearly values of the Stykkisholmur temperature record. The other stations have more of the same, i.e. changes to the original IMO data, but less obvious as at Stykkisholmur. Is this the aforementioned “quality control” at work ?

The idea and the database itself is excellent, necessary and commendable. But it should be done as a scientific task, not as another activist endeavor. Prof Ole Humlun – faculty of Mathematics and Natural Sciences at the University of Oslo, who also maintains the website – has recently started referring to the Hadcrut, NOAA and GISS world temperature series, all based partly on the GHCN database, as a lesser “quality class” compared to the 2 satellite datasets. In view of what I saw happening to the Icelandic temperature records, I would offer that he remains rather benign in naming them just 1 or 2 units less.

ducdorleans’ XL spread sheet

This entry was posted in Climate change, Political commentary and tagged , , , , . Bookmark the permalink.

21 Responses to Stykkisholmur (Iceland) temperatures: From reality to GHCN v1

  1. Flocard says:

    I am just wondering how many hours do the days last where you live (36h ? 48h ?) to allow you to do so much.

    • Euan Mearns says:

      Well, it was ducdorleans who researched and wrote this one. But I have to admit that the time I am spending on this venture is taking its toll and is unsustainable.

  2. Euan Mearns says:

    I have checked the UK temperatures for Feb 1929. It was indeed cold with mean night time temperature of -3˚C in Oxford (see table). The pattern of course is down to the jet stream, a feature well known to meteorologists and climatologists but only recently discovered by climate science.

    Duc’ is it possible to have a chart for Stykkisholmur that compares the original V0 with GHCN V3?

  3. Euan Mearns says:

    Thanks for this interesting expose Duc’. I’d draw everyone’s attention to the other stories in your second news clipping: 1) 50 collapse in Budapest scrambling for coal 2) Snow falling in Rome and 3) a sentry freezes to death 🙁

  4. carlos Leiro says:

    You understand that the data given by false theory of global warming? You should know the methods to obtain the data required to make an accurate analysis

    In Climate Change By Numbers, 75-minute special for BBC Four aims to remedy that, with a new perspective on the whole subject. Presented by three mathematicians – Dr Hannah Fry, Prof Norman Fenton and Prof David Spiegelhalter – it hones in on just three key numbers that clarify all the important questions around climate change. The stories behind these numbers involve an extraordinary cast of characters, almost all of who had nothing to do with climate change, but whose work is nevertheless critical to our understanding of the climate.

    The three numbers are:

    0.85 degrees (the amount of warming the planet has undergone since 1880)
    95% (the degree of certainty climate scientists have that at least half the recent warming is man-made),
    1 trillion tonnes (the total amount of carbon we can afford to burn – ever – in order to stay below ‘dangerous levels’ of climate change)
    Understanding how scientists came up with these three numbers gives a unique perspective on what we know about the past, present and future of our changing climate. But it also takes us on a journey through how science is conducted, to the very limits of what science can tell us.

  5. clivebest says:

    NCDC also do block adjustments to temperatures for all stations. This explains why V3C temperature values often appear to be systematically lower for early years. This is done using an automated algorithm which supposedly finds station moves – shifts in data. You can see the effect of this on Stykkisholmur here:

    The top graph in red shows the delta between V3U (uncorrected) and V3C (homogenised)

    Even more extreme are the changes made to VESTMANAEYJA in Iceland

    The red graph are changes made by NCDC and the green graph shows those made by CRU.

    A comparison between GHCN V1 and the latest CRUTEM4 shows just how much the early temperature data have been corrected downwards.

    • ducdorleans says:


      And that is just the adjustment from GHNC v3 to GHCN v1 !

      But we need to go further back … Iceland is simple, because they were so kind to put all of their data online … At all the other countries, we actually should go to the library of the various Met Offices, and check the temperatures in their various yearbooks against GHCN v1 …

      Most probably we would find something similar to all this: John Steinbeck’s Grapes of Wrath had some truth in it … Also the thirties were rather hot …

  6. Dave Rutledge says:


    Thanks for an informative post. Two criticisms one could make of the 2 degrees above pre-industrial temperature target that the bureaucrats aim at. First, it is pretty clear we do not know what the pre-industrial temperature was compared to today’s, and it appears that it is a constant temptation to bureaucrats to lower the official number. The lower they push down the pre-industrial temperature, the more power the bureaucrats can claim they need over our lives. Second, it is not clear to me that the pre-industrial temperature, whatever it was, was better than what we have today. I personally prefer the larger timber densities that we are seeing in American forests, and I think CO2, longer growing seasons, and even reactive nitrogen emissions from power plants are helping rather than hurting.


  7. Phil Chapman says:

    An anecdote about the perils of temperature data collection, from when I wintered in Antarctica, long ago:

    We had a met observer at Mawson named Helmut (he was an Austrian ski instructor who had joined our expedition because he wanted to try skiing in Antarctica, and was very disappointed that the katabatic wind swept almost all snow away, leaving bare dimpled blue ice). He was on duty one night near midwinter, when it was dark 24/7, and one of his tasks was to check the instruments every three hours in a Stevenson Screen up the hill behind Mawson.

    The temperature was -40C and the wind was above 80 knots, with dense blowing snow. Getting around required using a carabiner to tie yourself to the ropes strung around the camp. If you became disconnected, you might find yourself blown miles away out on the sea ice, and in any case you would be lost because visibility in the drift was about a meter.

    Helmut made his way laboriously up the hill, where he found that the screw was loose that secured the pen on the mechanical temperature recorder. He had not brought a screw driver, and he didn’t want to go get one, so he decided to freeze it in place. He foolishly touched the screw with his tongue, which of course froze instantly. There he was in the howling Antarctic night, attached by his tongue to the instrument. Pulling his tongue free would tear off a large chunk, but standing there for very long would mean death from exposure.

    Fortunately, he could undo the thumb screws holding down the instrument. So he staggered down the hill, holding the instrument in front of his face. He made it to the met hut, where he sat in front of the stove until his tongue thawed.

    Nobody wanted to embarrass Helmut about his folly in touching cold metal with his tongue, so I don’t think there is any official explanation for the incredible peak in the Mawson temperature record for that day.

  8. I’m wondering how these adjustments affect the extrapolation algorithms that GISS use to generate Arctic temperatures. Is big warming in the Arctic a function of big adjustments in the near-Arctic?

    • Euan Mearns says:

      Don’t know. I’m hoping that Ducdorelans comes back with an estimate of how these adjustments affect the temperature record. A few months back I did this analysis of Arctic temperatures in Scandinavia and Russia. It was just as warm in 1937. So when people make a song and dance about Arctic warming I don’t really know where they are coming from.

    • ducdorleans says:

      “adjustments” are a valid scientific method. Lubos Motl had a post about adjustments some months back, and was angry at anybody who was opposed to “any or some adjustments”. Trausti Jonsson, on his blog, explains the various reasons why some adjustments were needed to the recorded temperatures in Iceland. All the necessary, logical, scientific adjustments to the different Icalandic stations had already been made by the IMO.

      However, what we see here with the Stykkisholmur temperature in the GHCN v1 database can hardly be called “adjustments”. For a whole number of years, and months, there is no adjustment needed when simply transcribing the Stykkisholmur temperature from the various Danish and IMO Met yearbooks to the GHCN v1 database.

      And then, for a number of years, in some months, the sign is changed when doing that. The “adjustment” is equal to multiplying by -1.

      That sign change isn’t an adjustment, it is not a mistake, it is a deliberate effort to change something the perpetrators didn’t like. And what they didn’t like is obvious. They didn’t like “precedented”.

      But to answer the Bishop’s question. No, I don’t know … 🙂

    • lapogus says:

      Andrew, yes, you are at least right to ask this question. Intrigued by what Tony Heller and Clive Best were finding a few years ago, I remember looking at the adjustments to GISS data for stations around the North Atlantic (I think from v2 to v3) and it was noticeable that spurious adjustments (i.e. making the past colder) were usually of a greater magnitude to stations in Iceland, Jan Mayan, Bodo (Norway), than stations like Tiree and Stornoway. The former were >60N latitude, the latter 60N stations. I think have still got the GHCN v2 data (unadjusted and adjusted) for Reykjavik, Jan Mayen and Bodo,

      • ducdorleans says:

        Lapogus …

        “GHCN v2 (unadjusted)” doesn’t really exist … nor does “GHCN v1 (unadjusted)” …

        that is the point of this post ! .. 🙂

        of course, Iceland is just a tiny part of GHCN … and its influence on the “world temperature” is probably of Contadorian order of magnitude … but who knows what we would find when other REALLY original data would be compared to “GHCN (unadjusted)” …

    • manicbeancounter says:


      I would guess that these temperature adjustments have the opposite effect.
      First, there are the adjustments for outliers. The Stykkisholmur winter of 1929 was statistically an outlier. The flipping of the sign on the temperature would be due to someone guessing.
      Second, the temperature homogenisations conducted by GHCN (and then by GISTEMP on the same data) are pairwise comparisons over multiple runs. In the Arctic there is greater variation in average temperatures than elsewhere, so homogenisation can reduce the large swings. See for instance a graph of five year moving averages for Reykjavik and two Svarlbard stations.

  9. Just a quick reminder about a very common adjustment that riddles all the big datasets –
    How many times does a truth have to be told ? – UHI warming has been cemented into global temperature series by adjusting for steps outward from cities

  10. Pingback: Stykkisholmur (Iceland) temperatures: From reality to GHCN v1 | Climate Collections

  11. Hifast says:

    Excellent post and discussion folks!

Comments are closed.