the Air Vent

Because the world needs another opinion

Swiss Homogenization

Posted by Jeff Id on December 15, 2009

A guest post by email from a Swiss Slovak reader. I’ve not verified it but it seems to be in line with what we find elsewhere in the global temperature record.

Juraj V

Yesterday I read an article about Alpine glaciers melting even faster
in 40ties than today, while “temperatures were lower then” and less
cloudy sky was blamed.

I made a chart showing ratio of melting, stationary and advancing
Swiss glaciers vs AMO already some time ago:

Today I asked the Swiss glacier Institute for precip and temperature
data and they linked me to Swiss meteo web.
I downloaded temp history of their main stations, which contained raw
and homogenized temperature charts.

Striking was, that all stations after homogenization showed steeper
trend than raw data. There is one unpersuasive example of sensor moved
from city to airport cited. Average difference between raw and
homogenized trend was 2.09 times higher for the latter data. More, the
flatter was the raw trend, the bigger was the homogenization upward
adjustment, in one example rising the trend seven fold.

I downloaded collated Swiss homogenized annual temperatures, showing
1.13 deg C/century warming trend. Then I back-calculated raw trend by
adding a linear decrease to the homogenized data to get 0,53 deg
C/century trend (1.13/2.09) and compared it with two truly rural
stations – irish Armagh Observatory and Slovak Lomnicky peak

Click to expand

Swiss raw data and both rural records are in very good agreement,
while Swiss “homogenized” data are climbing twice the rate. Swiss raw
data are not treated for UHI as far as I read, when some of the
stations are in the cities or on airports.

46 Responses to “Swiss Homogenization”

  1. Third Party said

    Unprecedented and unexpected. 😉

  2. mrpkw said

    And worse then we thought !!!

  3. Dillon Allen said

    My “favorite” clarification of what nature meant to say is the next-to-last homogenization performed to make sure nature tells the right story. A 3-yr old who knows the difference between red and blue can spot this one.

    Climate Expert: Nature, you’re on in 20 minutes, please review the script and adjust your delivery to match the party line.
    Nature: garbled response
    Climate Expert: No, no, you won’t sound stupid for changing your story for no apparent reason at all. They won’t even notice. American Idol is on in the US and the rest of the world is watching soccer.

  4. Espen said

    Very interesting! I’ve seen the same looking at arctic stations – 60-70 years ago most places in the Arctic was just as warm as today. Nuuk, the capital of Greenland, was even warmer back then. It’s all extremely correlated to the AMO.

    If it’s possible to reconstruct a world temperature set without warmist bias, maybe we would see the same for world temperatures? That the current warm period isn’t really that much warmer than the previous period? I’m not a “denier”, I don’t object to there being a slight AGW. But I’m beginning to wonder if it’s almost insignificant compared to natural variation. If that’s the case, a doubling of CO2 may have less effect that we may wish for in a few years – if a new small ice age, maybe more severe than the previous one, is approaching.

  5. Anthony Watts said

    This is what you call a Swiss miss

  6. Hans Erren said

    check this
    A comparison of Basel, Vienna, Hohenpeissenberg, Uccle and De Bilt.

  7. dearieme said

    “emial from a Swiss reader”: I think there are holes in it.

  8. Jeff Id said

    #7 haha. I’m lisdexic with my typing of email.

  9. Ryan O said

    The other question I have is how much of the country data is pre-treated before it even makes it to GHCN. Hm.

  10. cogito said

    Funny, I was starting to compare Swiss figures raw/homogenized yesterday and came to the same conclusions. In ther explnation they say that the main reasons for homogenization were change of instruments and/or displacement. In these cases, wouldn’t you expect a step function?
    Look at the station SAE (Säntis), top row of last page in

    Säntis (Googel Earth: Saentis, Switzerland) is a peak at 2500 m altitude in a chain of mountains. The only building (which has been there for decades) is the top station of the cable car and a mountain restaurant. There is no way this station can be contaminated by urban heat, unless it sits on the roof of the restaurant. There is no way to move the station more than a few meters before it falls off. Yet, they have “homogenized” the temperatures.

  11. cogito said

    On comparing the station data for Saentis, Switzerland with GHCN, I noticed that the homogenized Swiss data
    is systematically lower (colder) by 0.2 to 0.4 °C than GHCN from 1864 to 1998, thereafter the difference becomes 0 from 1998 to 2009.

  12. Espen said

    #10: No urban heat on Säntis? Are you sure?

  13. Espen said

    #11: The “Säntis 2000” construction project (of the huge buildings seen in the image linked from my previous post) was finished in 1998.

  14. cogito said

    Some more pictures from the Saentis station:

    Weather station before 1935, on top of the hill:

    Telecom tower and old weather station

    This seems to be the most recent station

    Interesting story about this weather station:
    The restaurant at the top was built in 1864, the weather station in 1882. The weather station was manned throughout the year and daily readings were communicated to the base station in the valley . The cable car was built in 1935.

    In 1922, people in the valley were alarmed by the fact that no more weather data was transmitted for several days from the station. A crew was sent up to the station. They found the couple of weather guards murdered. To date, nobody knows what happened and who murdered the couple.

    @Rayon O: the GHCN data corresponds exaclty to the homogenized Swiss data except for the years before 1998 when GHCN is warmer by an offset of approx. 0.3 °C. A possible explanation: Switzerland “homogenized” (re-homogenized = cooled) the earlier data and GHCN didn’t synchronize the older values.

  15. cogito said

    Sorry: should be Ryan O.

  16. cogito said

    #13: 1998 corresponds to the time when the 0.3 offset between Swiss data and GHCN endet.

  17. cogito said

    Weather station on the Saentis and the murdered couple.

    Thermometer reading:

    Weather stations in Switzerland, the oldest was erected in 1838

    Talk about reliable temperature records …

  18. Ryan O said

    Cogito – which is, of course, another problem with GHCN. The NMSs already homogenize some of the data, and their procedures are not always well documented.

  19. cogito said

    Ryan O. I have sent a request for more details about the homogenization steps and reasons to Swiss Meteo.

  20. Juraj V. said

    Hi folks, here is the “Swiss reader”.. actually I am Slovak 😀

    Thanks to JeffId for posting that!

  21. Jeff Id said

    #20, Sorry about that.

  22. cogito said

    Some more findings on Saentis:
    As mentioned above #11, there is an offset between Swiss homogenized and GHCN raw of 0.2 to 0.4 °C.
    Going from GHCN raw to homogenized (again), there are variable offsets of up to 1.3 °C for all the years before 1978. These offsets seem to be random from month to month and year to year.
    In winter, GHCN homogenized is cooler, in summer it is warmer than GHCN raw.

    The offsets between a) Swiss homogenized and GHCN raw and b) GHCN raw and homogenized may, depending on their signs add or subtract from each other.

  23. Juraj V. said

    One station which puzzles me is Austrian Sonnblick on the mountain top, showing stronger warming than anything rural in Europe.

    Will try to search something about it on internet.

  24. […] Suiza: Swiss Homogenization […]

  25. Juraj V. said

    I have compared Swiss raw data with MSU tropospheric trend for 360*400km square covering Switzerland and there is only a very little difference. MSU data for Armagh grid and Lomnicky peak grid show no divergence whatsoever, confirming that MSU data show the same trend as rural ground stations. MSU grid data I got from KNMI website.

  26. cogito said

    Juraj, you mean Swiss homogenized? To my knowledge, there is no Swiss raw available. This
    is homogenized with a fairly strong warming trend.

  27. Juraj V. said

    Cogito, here are Swiss raw data, but only as charts with overall trends

    Click to access vergleichoriginalhomogen.pdf

    I back-calculated “raw” data somehow, see above.
    Raw data trends correspond with MSU for given area quite good, maybe showing a bit UHI?

  28. cogito said

    Juraj, Interesting. If raw trends correspond with MSU, the same cannot be the said for the locally homogenized data. Just looking at the station Sion (SIO –, we see a significant difference between raw (0.18) and homogenized (1.43). Since that’s what goes into GHCN, this would explain the divergence between surface temps and MSU, wouldn’t it.

  29. BillD said

    “I made a chart showing ratio of melting, stationary and advancing
    Swiss glaciers vs AMO already some time ago:”

    I am a little confused by this, since the author of the post claims to have made this graph, but URL on the top of the graph is attributed to the Swiss institute and NOAA.

    If we assume that the Swiss temperature data have not warmed and that solar insolation has declined, we have a hard time explaining the rapid melting of the glaciers during the past 25 years. We also have a hard time explaining data from Swiss lakes (search under Livingstone and Lake Zurich) which show substantial spring warming during the last 20 years.

  30. […]   […]

  31. Alexej Buergin said

    BillD: What is new about Jurai’s chart is that it is a combination of 2 existing charts.

  32. Alexej Buergin said

    I do not understand the use of the word “homogenization” in the context of “adjusting for wrong measurements”. Why did they not measure with old/new instruments or in the old/new place, compare, and make a step change? Would not “homogenization” mean the matching of data across the whole of Europe?

  33. cogito said

    Momogenization means adjusting for instrument, calibration, moving, replacement. But already mentioned, their homgenization leads to a steeper for all the stations, except one. There is no step function visible in theif graphs, and we dont have any raw data.

    About 3 weeks ago I asked Swiss Meteo for more details. They responded by asking me what exactly I wanted to have. I explained that I wanted to understand in detail how, why, and when individual station adjustments were made. I haven’t heard from them since, but given the Holiday Season, I’ll give them another week.
    I will report here what ever I might get.

  34. cogito said

    BillD: the glacier data is here

  35. cogito said

    BillD: and in digital form here

  36. Alexej Buergin said

    Cogito 33: Did you use this form: ?
    They promise an answer within 30 minutes.

  37. cogito said

    #36, I received an answer today.
    Translation. My remarks/interpretations in [].

    – The effect you have observed [the systematically increased trend of all the homogenized data] is a fact and has two reasons:
    First: The raw data provided until about 1900 was collected from {old fashioned?} thermometers which suffered from a [continuous?] slow contraction of the mercury containers. This had the effect of an increase of the measured value over the real value by 0.5 to 1 °C over several years or decades. This is corrected in our homogenization process [by deducting the difference = cooling the raw data].

    Second: At the end of the 70’s, many stations were changed to automatic versions. The measurement with ventilated thermometers resulted in a smaller influence by the direct sun exposure and hence lower average temperature readings, especially in summer [which requires increasing the data during homogenization].

    Besides this systematic corrections, there a station specific corrections which had the same effects and required a reduction of the measured values. e.g. in 1949 the station Zurich was moved from the old building of the faculty for physics at the Swiss Federal Institute of Technology close to down town [in fact: on a hill side above the city,-95.677068&sspn=45.197878,78.925781&ie=UTF8&hq=&hnear=ETH+Zurich&ll=47.378098,8.548093&spn=0.018976,0.038538&z=15&iwloc=A%5D to a new place at Zurichberg [at approx 2 km, at a higher elevation with a more rural surrounding].

    – In our view, the UHI effect has a minor role on our monthly data. Comparisons between rural and urban stations have not shown significantly different trends. A reason for this could be that our urban stations are all situated at the outskirts of the cities. Our homogenized data is therefore not corrected for UHI.

    – Our logs contain the station history but do not show which corrections were made in the data. Furthermore the station histories contain blanks and we can not be sure that all changes were recorded. This is particularly true for the early histories. For this reason, we also use statistical tests during homogenization by comparing to neighbored stations.
    The Saentis station was moved and corrected accordingly in 1882, 1975, 1978, and 1998 as well as temporarily in 2005. In 1978, the automatic measurement was introduced.

    – Regarding the data and the data quality used in the international data centers we only have limited information. MeteoSwiss feeds selected raw data [!] from selected stations to the WMO for use without restrictions [no non-disclosure agreement!]. The international data centers apply homogenization based on their own statistical evaluation without any feedback from us. They obtain no information about station histories.
    Their homogenization is necessary because only few member states provide homgenized data.
    It seems evident, that it would almost be impossible for the WMO to handle station histories from 192 member states [in almost as many different languages]

    Whether other data centers […] use our homogenized data is unknown to us, since the homogenized data is freely available from our web site.

    My comments: as mentioned above, in the case of Saentis, GHCN raw data is without a doubt identical to the Swiss homogenized data, with only a few minor differences for older data.
    GHCN homogenized data in absolute values has some resemblance with the Swiss homogenized data but contains systematic shifts in some periods, up to more than 1 °C in both directions.

  38. Alexej Buergin said

    According to this newspaper report
    there seemed to be a distince UHI-effect during the last ten years.

    So they only correct upwards. I wonder how much of the warming is real, and how much is adjustment.

  39. Denys Pierrehumbert said

    About 40% are adjustments from my evaluation, see the reference vergleichoriginalhomogen.pdf above.
    Sorry of my english, I am from the french part of Switzerland.

  40. cogito said

    “I am from the french part of Switzerland”. Me too.

  41. Alexej Buergin said

    So they believe that swiss meteorologists in the 19th century used the same thermometer for years and years and never thought of recalibrating them? Whereas in physics people were defining the “kilogram” (1889) in a way that is still in use today?

  42. Alexej Buergin said

    If Meteo Swiss feeds selected raw data to the WMO for use without restriction, there is no (financial or other) reason to hide it; they should publish it along with the adjusted data.
    Since “homogenization” is a far from settled science, they might even learn something from interested non-professionals.
    Not correcting for UHI e.g. (because the stations are at the outskirts) is nonsense when the population and the cities are growing. The city center may be “satiated” and the warming may happen mostly in the suburbs.

  43. Denys Pierrehumbert said

    My feeling is that the general move of thermometers is from center to periphery. During the live of a particular station, none homogeneization because of the “lake of UHI”. But UHI is like a wave from center to periphery passing over the station. When the station is moved, generally, it is to the outerside of the UHI wave. Then, the measures are falling and adjusted stepwise. Because “the UHI effect has a minor role”, the falling of temperautres is assumed due to a change of thermometer or any other effect. The more you move the station, the more you make warming. In french, I would say “homogénéisation à cliquet”. I think a part of the warming is due to this effect but I dont know how much.

    I studied the charts of Sion (SIO). I saw only stepwise homogeneization (1873 : +0.5, 1885 : -0.4, 1909 : +0.5, 1964 : +0.4, 1978 : +0.8 and 1982, 1983 with particulars absolutes values of +0.9, 1978 to 2008 is absolute +0.5). The sum 1864 to 2008 is +1.8, all these values are aproximatives. I guess 1909 and 1978 due to changes of thermometers but probably also of positions. Today, the station is in the lower possible elevation.

    Et bien le bonjour à Cogito.

  44. Alexej Buergin said

    Pierrehumbert #43: Using the example of Geneva, which Meteo Swiss cites, I assume that Cointrin was rural in 1960, and it was correct and necessary to adjust. But it looks as if the City has grown and is now “creeping” around the airport, which is probably more overbuilt and has more traffic, too (it is convenient, but principally wrong to situate stations for climate-measurements on airports). It may show the same temperature now as would be measured at the old observatory.

    (PS lake=lac, lack=défault)

  45. cogito said

    I’m living a few km from the Geneva airport. I’ll try to locate the station, but I suspect it will be either close to the runway or even on a building. There is not much rural space left around the airport, especially since the only green patches next to the runway are already on French territory.

  46. R.K. said

    I finally got the original data! It was a hard fight I can tell you. After lots of emails and phone calls (he called me when he did not know what to answer) I told him the following: the people at MeteoSchweiz think it is easier to legitimate that they hide the data than to legitimate the homogenization.

    The guy from MeteoSchweiz then forwarded me to a guy at the technical university in Zürich (ETH) who has acces to the databases of MeteoSchweiz. This guy was then so kind to give me original and homogenized data both starting in 1884.

    It is interesting to see the corrections, i.e. correction curves over the year, the step-wise changes, …

    I also asked for the documentation of the homogenization. The guy from MeteoSchweiz gave me the homogenization of one station (Zürich)

    What still I don’t understand is why the data is not provided by MeteoSchweiz for the public!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: