KNMI has been measuring the wrong temperature for years
WUWT reader Mike writes with this little bombshell on one of the world’s leading meteorological agencies. It seems they can’t get their thermometer siting correct which resulted in a bias to the record. Hmmm. Where have we heard this before? The newspaper “AD” in the Netherlands has picked up the issue with two separate stories.Mike writes:
Dear Anthony,
I left this on the “tips” thread on WUWT, but since it is also relevant to surface stations, I felt you should hear of it directly. It probably deserves a whole story on WUWT.
As you probably already know, KNMI De Bilt is the only station in the Netherlands used for GISTEMP. The nearest long-term station is in a suburb of Brussels, hence is undoubtably UHI-polluted. De Bilt is the only long record stn in NL & within 150km in any direction would be a useful correction.
Two stories caught my eye in the Dutch papers today about a 0.5-degree error in the De Bilt record which was miraculously corrected this summer with a station move of 200 m without anyone being told of it. Here are the links to and my translations of the articles.
Mike’s translations of the newspaper stories are below, I’ve added relevant graphics. – Anthony
The instrument stood too close to a line of trees, due to which on average half a degree (Celsius) too high was measured.
After discovery of the fault the thermometer was moved to an open spot on the measurement field before last summer, the KNMI has confirmed. Due to the change the average measured temperature fell half a degree. This measurement should be reliable.
The mistake resulted in that the KNMI has announced more “official” summery and tropical days than there were in reality. According to the Institute, the defect has not or hardly influenced the scientific discussion on climate change, because researchers use the data from a large number of weather stations. SUZANNE DOCHTER
Above: GISS Temperature plot for De Bilt KNMI – notice the step function. Click for source data.
Here’s a picture and metadata for De Bilt, direct from KNMI. While I can’t be certain, this photo appears to be after the move:
Checking some nearby stations in GISS, click for source data:
The GISS plot for Maastricht Airport:
One whole data point? Why does GISS keep a station in the database with only one data point?
UPDATE: Well if GISS can’t find the data for Maastricht Airport, everybody else can, and damn quick:
See Weather Underground for current conditions.
And this website, tutiempo.net , has the complete climatic data set back to 1949.
http://www.tutiempo.net/en/Climate/Maastricht_Airport_Zuid_Limburg/63800.htm
Call out to GISS: Hey Gavin, as an American Taxpayer that funds your work, I request you take a moment from moderating realclimate.org and put some work into updating this record.
Here’s the next closest station, Essen, according to GISS a city of 7.5 million – doesn’t look much like the KNMI record:
2nd story —
KNMI has been deaf to criticism for years
WAGENINGEN – weather Institute KNMI has been deaf to years of criticism from competitor Meteo Consult of its temperature measurements in De Bilt
Weather specialists from the Wageningen-based Meteo Consult have been expressing their distrust for years, because the KNMI figures in De Bilt were always a bit warmer than in Cabau, 16 km away, where there is also a KNMI thermometer. The position of both places could, according to Meteo Consult, not explain the temperature difference of on average half a degree (Celsius). It was also not taken into account that De Bilt is located in a more built-up, and probably therefore warmer, surroundings than Cabau, near IJsselstein.
The meteorologists from Wageningen discovered this summer to their amazement that the temperature difference between both places in the KNMI figures had more or less disappeared. On enquiring of the De Bilt employees, it appeared that the thermometer had been moved. Since the intervention, the measurements from De Bilt show not 1/2°, but on average just 2 hundredths of a degree warmer than Cabauw, according to the spokesman of Meteo Consult.
This summer it appeared that the temperature difference was suddenly resolved. Again discussions blazed between the weather specialists and it was decided to closely compare the measurements between Bilt and Cabauw. “It was thus discovered that last summer in De Bilt was still 1/2 degree warmer and this year there was just a difference of 0.02 degree Celsius”, explained a spokesman of Meteo Consult.
The organisation decided to call the KNMI and heard that the “weather cabin” [translation: Stevenson Screen], in which the thermometer is located, had been moved. According to the KNMI the measuring instrument stood too close to a row of trees. Because the trees continued to get taller, the wind began to influence the temperature measurements too much. Now the “weather cabin” has been moved 200 m away, to a more open spot on the measurement field of De Bilt. KNMI employee Cees Molenaars cannot say how much influence the old placement of the thermometer has had on weather reports. “We must investigate that. We only regret is that we did not keep Meteo Consult and other parties informed of the movement.”
The thermometer of De Bilt is the official measurement used for determining heatwaves, cold waves, and summery days. To speak of a heatwave it must be at least 25°C released 5 days. Also it must be warm than 30° for 3 days. At 25° one can talk about a summery day.
With a cold wave, freezing temperatures must be measured for 5 adjoining days at De Bilt, with also 3 days with a hard frost. “The differences in minimum temperature between de Bilt and Cabauw were much smaller,” said the spokesman of Meteo Consult. “The chance that a cold wave is missed, is thus smaller.”
The thermometer in De Bilt has less influence on KNMI weather predictions. These are performed on the basis of the data of tens of measurement stations. Further, for scientific purposes, such as climate change research, the central Dutch temperature was brought to life long ago. For this, data from various stations is used [NOT TRUE — GISTEMP ONLY USES DE BILT!]. Meteo Consult are above all happy that the riddle has been solved. For fun they have also calculated what an extra half degree in De Bilt would have meant for this summer: 5 extra summery days and 2 tropical ones.
====================
Coincidentally, I’ve been conversing with Jos de Laat of KNMI, the Dutch Meteorological Institute who offered some scans of weather station siting specifications from the World Meteorological Institute (WMO)
he writes:
OK then, you can find the first part of the report here (~ 1 Mb):
http://www.knmi.nl/~laatdej/TMP/WMO488.pdf
Especially the beginning of part 3 is relevant, I guess. Because of document size considerations for now I only scanned up to paragraph 3.1.2.1.7 (after paragraph 3.1.2.1.7 the description of requirements for measuring on other locations like sea and the free troposphere starts).
Descriptions of sensor and siting requirements are also available online (see below) …
… but they are more formal and largely based on WMO report 488, which contains some interesting quotes that are not present in later reports. The online reports also refer to the report below, which unfortunately I was not able to locate either online nor in our library.
World Meteorological Organization, 1993a: Siting and Exposure of Meteorological Instruments (J. Ehinger). Instruments and Observing Methods Report No. 55, WMO/TD-No. 589, Geneva.
These specs are worth a read, because they show that quite a lot of thought and analysis went info choosing the specs.
As for the 100 feet cited by the NWS on this page: http://www.nws.noaa.gov/om/coop/standard.htm
I suspect its a round off of 30.48 m where 30 meters is the minimum distance to an artificial heat source cited for a Class 2 climate site as defined by the specs used in the Climate Reference Network (CRN) which has a French lineage, and likely traces back to WMO.
It seems that no matter where you look, meteorological agencies can’t follow siting specifications.
Steve McIntyre of Climate Audit has more on De Bilt and the adjustments that are being applied there:
http://www.climateaudit.org/?p=1650
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
When I checked the temperature-data from De Bilt I found that the climate changes more during the night than during the day.
For the period 1901 – 1920 the mnimumtemperature was 5,0 degress C.
Between 2001 – 2008 average minima were 7,3 degrees C.
Minimumtemperatures have increased by 2,3 degrees Celsius.
The maximumtemperature increased from 13,3 to 14,7 degrees Celsius.
A smaller increase of 1,4 degrees.
When I asked a question about UHI-effect on the minimumtemperatuur, the KNMI replied that the effect was probably not significant. The difference between day- and nighttemperature can also be caused by other factors.
The data are summarized in: http://img200.imageshack.us/img200/1009/1250846805333.gif
Ben Lankamp (08:10:50) :
“No adjustments were applied to the official (non-homogenized) KNMI record. The GHCN contains outdated or plain wrong records of De Bilt, of which I have already notified NOAA through the proper channels. Hopefully an update will be made to the GHCN soon.”
The record in http://www.mscha.org/knmi/summer.cgi?station=260&sort=year&month=-1&columns=more does not contain the step change in 1951 which is obvious in the GISS record. And this step change is discussed in http://www.knmi.nl/publications/fulltexts/hisklim7.pdf
To me it looks like the unadjusted GISS record is unadjusted for the 1951 step change because of the station relocation, while the record in http://www.mscha.org/knmi/summer.cgi?station=260&sort=year&month=-1&columns=more is indeed adjusted.
Do you have a direct link to the non-homogenized record you mention above?
…but what if the step change in 1951 was real?
Uccle actually shows a very similar step change! – but on the other hand, Uccle seems to have gotten an additional station in 1951 too, so they may have introduced a similar error. However, right on the other side of the canal lies Gorleston, which ALSO had a big temperature drop in 1951:
http://data.giss.nasa.gov/cgi-bin/gistemp/gistemp_station.py?id=651034960010&data_set=0&num_neighbors=1
The 1950 step is a well known WMO artifact
Below is the reported well documented move in De Bilt:
http://members.casema.nl/errenwijlens/co2/debiltlocations.jpg
Compare adjustments to GHCN data by GIS, van Engelen & Nellesein (KNMI) and myself.
http://members.casema.nl/errenwijlens/co2/debilthomogenisations.gif
http://members.casema.nl/errenwijlens/co2/homogen.htm
and this interactive google map
http://maps.google.com/maps/ms?ie=UTF8&hl=en&msa=0&msid=108455475154621623933.0000011301001df084755&ll=52.100488,5.177125&spn=0.002755,0.007081&t=h&z=17&om=1
Hans Erren: Thank you for the links! If the 50/51 step is a well known WMO artifact, I’m surprised that it’s not removed in the homogenized GISS records.
Regarding the UHI adjustments in the article you refer to, I’m surprised that one uses remote Fichtelberg to adjust dutch data – the climate of Fichtelberg in Sachsen is obviously very different from the dutch climate, so don’t you introduce a lot of noise that way?
I would have thought that the US and Western Europe would be where the best measurements of Temperature are made. Not so good, as we now know.
I wonder what Anthony would find if he had the time to look at European WX-stations. And I prefer not to think about what he might find in some other places.
Darwin, at the North centre of the Australia coast, is a bit important when examining El Ninos. The land temperature at Darwin over about 100 years can be found on the Internet and on Bureau of Meteorology products, as adjusted by various parties.
Here is one version (I have applied minor in fills to missing data here and there, not significant in showing the picture, but not reproducible for detailed maths).
Given that all of the data originate from one source (the earliest being labelled BOM CD 1993, it was not a CD but a dowload) it is fair amazing how much variation there is. Plotted are annual averages from daily records.
If you wanted to correlate a climate property with temperature, which curve would you select and why?
http://i260.photobucket.com/albums/ii14/sherro_2008/DARWIN_SPAGHETTI_2.jpg?t=1253928389
to Clint Hotvedt
[snip]
Moving the sensor out from under the trees and into an open area would not have made the night time temperatures warmer. The trees, like concrete and building walls, would be a heat souce at night and would have warmed the night time readings. That’s why there is a urban heat island effect. You obviously need to stay in school much longer.
Giss Data not homogenise shows much the same as the series 1,2,3,4 from 1882 – 2009
The homogenised is only from 1963 – 2009 but again follows 1,2,3,4 (1963 to 1970 are similar temps to 2000 to 2005)
So where have you obtained series 5?
In the Smithsonian Miscellaneous Collections Volume 79 page 559 with corrections on page n43, there are records for Utrecht-DeBilt going from 1849-1900. Later volumes contain the more recent records. I expect records like these may be the “outdated” or incorrect records, or the source of the multiple versions.
Could one of the knowledgeable commenters from the Netherlands explain how records from that long ago are determined to be outdated or incorrect? I think this question ties in with requests for raw data recently done at Climate Audit, as I believe GISS and its predecessor agencies as well as Hadley Center all used these Smithsonian Collections as their first resources when building their climate history records.
Anthony,
Several people here have pointed out the errors in the newspaper articles. So for the sake of presenting balanced and objective information, it would be fair to update your post and make clear that the two newspaper articles you quoted are erroneous.
At the very least it should be noted that there is no basis for the statement that there was a “on average half a degree error” at De Bilt. This statement apparently was made up by a reporter who wanted to write a juicy story, and then other reporters eagerly copied it without checking it out themselves.
REPLY: Actually I have something better – the internal report prepared by De bilt that shows the results of a running field experiment on this issue, which I’ll publish soon. – Anthony
Espen (14:40:12) :
The climate in Western Europe is remarkably similar as this graph shows from observation points 1400 km apart
http://members.casema.nl/errenwijlens/co2/europe.htm
I also used these stations to identify the inhomogeneity in the Hohenheim temperature data
http://www.climateaudit.org/?p=7134#comment-357111
So yes Fichtelberg can be used as a homogeneitytest. Of course if stations are remote, small inhomogeneities can disappear in the noise.
a reader (05:59:18) :
I obtained the long GHCN series from the GISS website, before James Hansen decided to truncate all records at 1880
Full record is still available at KNMI
http://www.knmi.nl/klimatologie/daggegevens/antieke_wrn/labrijn_ea.zip
A van Engelen and Nellestijn, JW, 1996, Monthly, seasonal and annual means of air temperature in tenths of centigrades in De Bilt, Netherlands, 1706-1995. KNMI report from the Climatological Services Branch
this maps shows the sites that are used to make the series
http://maps.google.com/maps/ms?ie=UTF8&hl=en&msa=0&msid=108455475154621623933.0000011301001df084755
Original source:
A. Labrijn, 1945, Het klimaat van Nederland gedurende de laatste twee en een halve eeuw, Mededelingen en Verhandelingen nr. 49, KNMI publ. 102. (PhD dissertation Utrecht University)
Espen (12:17:30):
Data is available from KNMI in comma-separated format: http://www.knmi.nl/klimatologie/maandgegevens/datafiles/mndgeg_260_tg.txt
Notification from KNMI with regard to this record: due to possible relocations and changes in observation methodes these records of monthly values are not homogenic.
There is no step change around 1950 in this record, so possibly the 1950 relocation, which Hans already described, is the -only- thing corrected for in the record above. For the rest it is non-homogenized data (according to KNMI). The official homogenized version of the De Bilt record can be found here:
http://climexp.knmi.nl/data/temp_De_Bilt_hom.dat
Source: A.P. van Ulden, The Construction of a Central Netherlands Temperature, KNMI-WR 2009-03, available online: http://www.knmi.nl/publicaties/fulltexts/CNT.pdf
Re: Alexej Buergin (12:56:09) :
Is our RR Kampen not a Dutch meteorologist? Where is he when he is needed? Can he be prodded to come out of the woods?
(I just hope he does not work at KNMI)
and
Alexej Buergin (07:59:42) :
Do I understand correctly:
RR Kampen IS a emloyee of KNMI?
And there are THREE different set of records by KNMI at GISS? Why?
And summer 1947 WAS warmer in Holland than anything recent?
—
I am no employee of KNMI or any other weather/climate institute, although in 2006 I was close to a job of operational meteorologist at WeatherNews. But I’m into software quality assurance for a company not at all related to weather/climate. Nevertheless I’m ‘into weather and climate’ since I was 11, 31 years ago, have some years university on this subject until I went into mathematics to study (‘chaotic’) dynamic systems, am an active participant in the Dutch meteorological community.
For the sets of de Bilt records at GISS, see explanations by Ben Lankamp.
1947 is still the hottest summer in the Dutch record, though individual summer months July and August carry high records of very recent years.