Many readers are familiar with the Central England Temperature Record (CET), now Frank Lansner investigates the Central Netherlands Temperature Record (CNT). Long, but enlightening. – Anthony
Guest post by Frank Lansner
Instead the KNMI has decided only to make available their “CNT”, the Central Netherland Temperature index.
KNMI do make two coastal temperature series available before 1951 (Vlissingen and De Kooy) and then some strongly adjusted temperature series (De Bilt and Maastricht). Finally the Northern Eelde Station that has not been adjusted has been made publically available, fortunately.
To make the “CNT”, KNMI use strongly adjusted versions of their temperature data, let’s start out with Maastricht:
Fig2
Before 1950 close to one full Kelvin of heat has been taken out of data. The strong dive in temperature 1950 has been removed.
Fig3.
Exactly the same occurs for the De Bilt station, the temperature dive in 1950 has been removed.
Fig4
Here the combined adjustments of Maastricht and De Bilt data. The “Central Netherlands Temperature” index is shown here too. It is almost identical to the adjusted De Bilt and Maastricht data.
Fig5.
The raw temperature data from Bruxelles station Uccle obviously confirms the raw versions of De Bilt and Maastricht. Perhaps slightly more urban heat warm trend can be spotted in the Uccle data. But check this:
Fig6.
It does seem that the other Bruxelles station “Bruxelles National” has less warm trend than the Uccle station, suggesting a little UHI in the Uccle dataset?
Fig7.
Let’s move on, the Luxemburg station is available from Crutem3, and the dive in temperature from the late 1940´ies to 1956 yet again is confirmed.
Imagine that KNMI was correct in their adjustments and just by coincidence we have De Bilt, Maastricht, Uccle and Luxemburg showing a HUGE freak error simultaneously in 1950.
Would it take a miracle at this point for KNMI to be correct?
Well, let’s move on.
Fig8.
From GHCN we have Frankfurt-Wiesbaden temperature data again the warmer pre-1950 data are confirmed. In addition the warm peak 1957-62 is stronger in central Germany.
In my previous article on this matter I analysed a row of German stations from raw GHCN V2, check link in the end of the article.
Fig9.
On the site “Tutiempo” you can normally find adjusted data but I checked out this “Ypenburg” station. It appears that in the period approx. 1936-55 there was an Airport in Den Haag called Ypenburg. Tutiempo for some reason holds just a little sequence of apparently raw data from Den Haag Airport:
Fig10.
The Ypenburg station is around 10 km from the coast. The “choice” of years available from Ypenburg is perfect: We have once again confirmed the large dive in temperatures 1949-1956 in Southern Holland.
How about Paris data? They must be available? For some reason NO raw temperature data is available before 1951 from all Northern France. But Paris then?
Fig11.
Obviously we should expect some UHI for a Paris station, but none the less, the Paris Orly data – (although not showing 1956) do confirm the warmer pre 1950 temperatures compared to the 1970-80 level. Also Paris do not support KNMI and their “CNT”.
Fig12.
When I first looked at Paris “raw” GHCN data I was surprised because the year 1949 showed a large temperature dive unlike the other stations in the area. However, the 1949-dive in GHCN Paris L Bourget was contradicted not only by all other stations in the area, but also by the other Paris dataset, Paris Orly, here taken from Tutiempo. Thus, all Northern French raw data are eliminated in GHCN before I 1951, and then the only dataset with a 1949 peak happens to have the data point lowered 2 K.
Fig13.
Before returning to the Netherlands, Finally one more German station from raw GHCN V2, Dresden (for more German and Czhech stations, see the link 2 in the end of the article).
Gemert:
Yet another Dutch temperature station seems to behave wrongly according to KNMI, and then the error happens to take place in 1950, so that KNMI have to lower the pre-1950 temperatures:
Fig14.
This is how KNMI illustrates the corrections done to the Gemert station. They compare to a reference dataset I believe is not public, but that is likely to resemble the CNT.
KNMI explains the changes to the Gemert station:
“
Gemert had a large break in October
1949, when the station was renovated. In the period 1906_
1949 the record shows a significant positive trend relative to
the reference stations Oudenbosch, De Bilt and Winterswijk.
This trend was likely to have been due to a gradual growth of
the vegetation at this station until the re-instalment in 1949
“
So, the increasing divergence 1906-1950 with (already adjusted!) De Bilt and more is due to plant growth, and the change 1949-50 is then due to re-instalment, KNMI says.
Fig 15.
Same scenario, this time its Uccle divergence from the KNMI “homogenized” De Bilt data set.
So the Uccle increase in divergence 1906-1950 to the “homogenized” KNMI De Bilt data is also plant growth then? And also a re-instalment in 1949-50 in Bruxelles, perhaps?
But the increase in divergence is even faster 1880-1906 – So plants grew even faster before 1906?
Fig 16. From figure 5 of [1].
1) Its definitely possible that I misunderstand this figure, but as I understand, it shows the divergence between individual stations and then a reference trend based on data from Netherlands? If so, how come al stations show a positive divergence 1940-50? Should a valid reference not be made so that it resembled actual temperatures of Netherlands as much as possible?
2) The Maastricht and De Bilt stations only differ from this reference with around 0,15 K.
As the difference between raw Maastricht/De Bilt versus CNT is around 10 times as much, this suggests that this figure actually show divergence between a reference and already ADJUSTED temperature sets. I’m not sure what scientific value this has.
3) The Eelde divergence is shown lower than the De Kooy divergence. As we will see later, Eelde is roughly 0,5K warmer than for De Kooy before 1950.
So how come they can show De Kooy with a warmer divergence than Eelde?
Again it seems that several data sets have been adjusted before showing divergences in the above illustration.
One more note: The illustration do not show data points for Leeuwarden 1949-55?
Fig17.
Many Tutiempo temperature series I have been able to test against raw data appear to be warm adjusted. But the point here of mentioning Leeuwarden none the less is that we from Tutiempo learn that this data set do exist at least from 1949.
So, why did KNMI only use data from 1955 in their illustration?
Eelde:
Fig18.
In Northern Holland we also have the KNMI data for Eelde and it can be proven unadjusted against NACD V1. Eelde data resembles Leeuwarden data from Tutiempo, and thus the Tutiempo Leeuwarden dataset also appear unadjusted.
In comparison with the previously shown apparently raw datasets, this Northern region with the Eelde and Leeuwarden stations appears to have had a slightly colder period 1930-50 but still around 0,5-0,6 K warmer than the “CNT”.
Coastal temperature stations of the Netherlands.
Fig19
When examining data we know is raw (or have a reason to believe is raw), then only the coastal stations Vlissingen and De Kooy show similarity with the “CNT” temperature trend before 1950.
CNT appear to be the “Coastal Netherlands Temperatures” rather than the “Central Netherlands temperatures”.
Fig20.
We can now estimate coastal and Non-coastal temperature trends for Benelux based on coastal stations De Kooy and Vlissingen and non-coastal stations Uccle, Luxemburg Airport, Eelde, Maastricht and De Bilt. All are raw datasets. Obviously stations Uccle and Maastricht are likely to include some urban heat.
Fig21.
NE USA coastal and Non-coastal temperature trends. Somewhat similar to the Benelux data sets. (taken from link 3: From RUTI Coastal temperature stations.)
Closing comment:
Yes, here and there I cannot be 100% sure which stations are adjusted and which are not.
The issue here is that raw data from KNMI is not just freely available, that would be a lot easier. But since this is not the case, I find it better to try to give the best estimate possible.
Bonus info.
Fig22.
The distance ocean air influence over land is illustrated her for SW Jutland [4]. Most of the ocean effect disappears just around 5 km from the coast (depending on topography also). Therefore the poorest stations to use for land temperature estimation are the coastal stations. However, many hundred kilometres from the coast, still the coastal trends can dominate temperatures on hills and mountains, and sometimes valley stations just downstream from larger mountains.
Links:
1 The creation of a Central Netherlands Temperature, KNMI:
www.knmi.nl/publications/fulltexts/CNT.pdf
2 NW Europe and De Bilt (more details on German stations from raw GHCN V2)
http://hidethedecline.eu/pages/ruti/europe/nw-europe-and-de-bilt.php
3 More on Coastal temperatures.
http://hidethedecline.eu/pages/ruti/coastal-temperature-stations.php
4
http://img.kb.dk/tidsskriftdk/pdf/gto/gto_0047-PDF/gto_0047_69738.pdf
If you have the time, please cut and paste the below temperature data while they are online…
http://www.tutiempo.net/en/Climate/Brize_Norton/36490.htm
http://www.tutiempo.net/en/Climate/Prestwick_Airport/31350.htm
http://www.tutiempo.net/en/Climate/Bremen/102240.htm
http://www.tutiempo.net/en/Climate/Hamburg-Fuhlsbuettel/101470.htm
http://www.tutiempo.net/en/Climate/Hannover/103380.htm
http://www.tutiempo.net/en/Climate/Nuernberg/107630.htm
http://www.tutiempo.net/en/Climate/Koeln_Bonn/105130.htm
http://www.tutiempo.net/en/Climate/Paris-Orly/71490.htm
http://www.tutiempo.net/en/Climate/Szczecin/122050.htm
http://www.tutiempo.net/en/Climate/Gdansk-Rebiechowo/121500.htm
http://www.tutiempo.net/en/Climate/Gdansk-Rebiechowo/121500.htm
http://www.tutiempo.net/en/Climate/Koebenhavn_Kastrup/61800.htm
http://www.tutiempo.net/en/Climate/Bruselas_Bruxelles_National/64510.htm
http://www.tutiempo.net/en/Climate/Leeuwarden/62700.htm
http://www.tutiempo.net/en/Climate/YPENBURG_NAFB/62000.htm
http://www.tutiempo.net/en/Climate/Guernsey_Airport/38940.htm
PLEASE GIVE ME A TIP IF YOU HAVE KNOWLEDGE ON
RAW TEMPERATURE DATA IN WRITINGS ANYWHERE IN THE WORLD
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Numbers: http://www.knmi.nl/klimatologie/daggegevens/download.html .
And sorry mates, there is no dive in 1950. The years 1950 and 1951 were meteorologically uninteresting in this area.
Half of the Netherlands where flooded in 1953.
http://en.wikipedia.org/wiki/North_Sea_flood_of_1953
Should have had some effect on the inland temps.
And SRJ..
You did notice that the explanation for changing data in Gemert was quite different than the story you got?
But all accross Holland, Belgium, Luxemburg, Germany, France, Denmark the temperature drop around 1947-56 is wrong, but for different simulktaneous events? We also have the old DDR represented, I guess they co-worked with Holland measuring city temperatures in 1950 ?
Frank, you may want to cross check with Coolhansnl
http://members.casema.nl/errenwijlens/co2/index.html
A few remarks from me as a Dutchman:
@ur momisugly Philip Bradley: absolutely valid remark about the influence the Flevopolders may have on the climate, causing De Bilt to be effectively further away from water. Technically speaking not from the sea, as the former Zuiderzee (“south sea”) had already become a large sweetwater lake since the closure of de Afsluitdijk in 1932.
Nevertheless, de Flevopolders added substantial land surface.
But the Flevopolders only became dry in 1957 (Oostelijk Flevoland, easterly part) and 1968 (Zuidelijk Flevoland, southerly part) respectively. I happen to live in the easterly part (Lelystad), 4 meters below the sea level.
Lelystad by the way set a record low temparture this winter beginning of February at minus 22,9 celsius. It was the lowest temperature recorded for the last 27 years in the Netherlands. Of course this doesn’t say much about averages and trends, but at least this “children won’t know what snow is” and “we won’t ever be able to iceskate outside anymore” has been thoroughly debunked. It is also encouraging and funny that the large public has remembered these claims and many now deride them.
@ur momisuglyAtomic Hairdryer: altitude is largely irrelevant in the Netherlands, with the exception of Maastricht.
General remark: the people at KNMI are (unfortunately) known as fierce warmists. Luckily we have some other voices as well in the Netherlands such as climategate.nl and scientist Mr Bas van Geel.
Is there anywhere where temperature data hasn’t been scammed in one way or another ?
This rings some bells – we’ve discussed some of these issues before. In a thread 2,5 years ago, Hans Erren told me that “The 1950 step is a well known WMO artifact”. I’m still not sure what this artifact actually is – change in measurements standards? Anthony, do you know? Or maybe Hans will show up and explain?
Quote beginning
Frank Lansner says:
”
May 18, 2012 at 3:04 am
SRJ:
After 1950, measurements for all 24 hours are sent from de Bilt to GHCN. Thus giving a much better representation of the daily cycle.
That means that prior to 1950 the daily means in GHCN v2 are too high.
No.. any change in data could just as well result in a cold adjustmens as a warm adjustment.
Besides, ITs not only GHCN raw V2 that has the original De Bilt.
If this is a general Dutch issue why are not the coastal stations changed in 1950 like De bilt?
And howcome the 1949-56 temperature dive are seen in the near by countries?
And why are we not allowed to see any data (except strongly urban) for Belgium etc?
K.R. Frank ”
Quote ending
This particular change I am describing is introducing a warm bias as I demonstrated to you in klimadebat.dk. The example with synthetic data is shown here:
http://www.klimadebat.dk/forum/vedhaeftninger/temp.jpg
I am planning to make an example using the hourly data for de Bilt from KNMI’s homepage.
I do not think this is a general dutch issue since it is only described for de Bilt in the article by G. van der Schrier. But it is an issue for de Bilt, and it is introducing a warm bias and therefore, data from GHCN v2 for de Bilt is not representing the correct de Bilt temperature.
cRR Kampen says:
“Numbers: http://www.knmi.nl/klimatologie/daggegevens/download.html .
And sorry mates, there is no dive in 1950. The years 1950 and 1951 were meteorologically uninteresting in this area.”
Kampen, Please examine you own link:
Start dates are later than 1950 excpet for coastal stations and the strongly adjusted De Bilt and Maastricht.
Only Eelde – exactly as said in the article – in the North of Holland is RAW an UNADJUSTED available before 1951, and as shown Eelde do have 0,5 K more dive after 1949 than the CNT.
It is exactly as I say: There are no RAW temperature data available for the Dutch people before 1950 except for coastal stations and Eelde. See the first figure in the article.
leftturnandre says:
“Frank, you may want to cross check with Coolhansnl
http://members.casema.nl/errenwijlens/co2/index.html
”
Thankyou so much, I will check it out from A ti Z.
> also you cannot compare raw absolute temperatures unless you correct
> for altitude differences and latitude differences in stations since both latitude
> and altitude explain a majority of the variance in temperature between stations.
This is fair enough if you’re trying to determine the temperature at one particular place that’s at a different position than one of your thermometers.
But if you want to determine a trend, nothing should be adjusted. It should all just
be averaged to get a true, unbiased result.
“”SRJ”
You write: “I do not think this is a general dutch issue”
This is more true than you might realise.
The non-coastal drop from the 1930-50 level to the 1960-80 level can be traced basically from Morocco to Russia.
its only THE SHAPE of data that differs a little , but the KNMI´s Rambo-like data changes removes the whole trend difference 1930-50 versus 1960 that is evident also From Ireland to Isreal!!
Come by one day and I will show you 😉
K.R. Frank
Frank Lansner at May 18, 2012 at 4:35 am
I’m sorry, I’m Dutch and into the met as of over thirty years. I know more than three centuries of climate and climate history in this country practically by heart and I know it utterly, not only by raw data but by all those factors influenced by weather and climate conditions.
One remark to be made: where you measure in this country is not too relevant. It is flat and it is very small. A trend in Maastricht (my birth city by the way) will be found likewise, with at most a couple of tenths of degrees difference, everywhere in the country. Recent warming by the way was strongest by that small margin in the far northeast of the country, which has neither urbanized nor seen land win over de past half to whole century. The data voor de Bilt, and Zwanenburg if you wish to go back a while, suffice.
By the way, Eelde/Groningen is not coastal like De Kooy is (for a meteorological reason). Also, there are more numbers, e.g. http://www.knmi.nl/klimatologie/daggegevens/antieke_wrn/index.html .
Those who like to try and find out whether meteo-wise the years 1950 or 1951 were anything special, as compared to e.g. 1947, 1956, 1963 or 1979, should take a Sunday afternoon to leaf through an archive like this: http://www.wetterzentrale.de/topkarten/fsreaeur.html .
Frank Lansner at 5:12 am, why not show now?
I’ve noted no little disaffection with the recent (and not so recent) ‘adjustments’ of temperature data sets, and the resultant (further) distrust of climate science. Some comments here on WottsU for example:
“Has there ever been an example of data being “adjusted” to make AGW less? Every time I see an “adjustment”, it makes recent temperatures hotter and past temperatures colder.”
“It would be a joke if not that so many billions of people are hurt or killed by these charlatans.”
“The Team simply made the whole thing up from the beginning.”
“WHAT A SCAM! What proof that NOAA is HIDING INFORMATION THEY DON’T WANT US TO HAVE!”
I could go on, but I think you get my drift.
The problem is that using this adjusted data leaves ‘sceptics’ playing off the back foot (“Look at GISS/CRU etc. graph! Now the World is warming up again”). The more adjustments, the futher sceptics are being painted into a corner! How can sceptics argue that the world is cooling while ‘climate science’ is shifting the goal posts all over the place (more precisely to where it’s warmer).
The answer is that the sceptic movement must completely cut its ties with the manipulated graphs of the (quote) charlatans. A completely new, true, uncontaminated data set for Global Temperatures must be produced by trustworthy people. Legitimate adjustments can be made to the raw data, of course. This website has proven that there are more than enough absolute amateurs who are up to the task. Frank L. appears a likely candidate; Steve “the hero of the 21st century” McI., too.
Let’s face it, sceptics are on a hiding to nothing at the moment. There’s a desperate need for some proper graphs which go the right way. How’s about starting with Global Temps from the late 1800’s, then a proper graph of the satellite record. If everything goes well, move on to Sea Ice and the other manipulated graphs….
Good idea? Or what?
The Bilt was replaced in 1950. It cooled the record by 1 C. It explaines the 1950 dip in raw data, but it does not explain the other stations.
Jeroen says:
“The Bilt was replaced in 1950. It cooled the record by 1 C. It explaines the 1950 dip in raw data, but it does not explain the other stations.”
Im happy we agree on this.
One thing i would like to mention though: KNMI suddenly 60 years after 1950 assumes that a local station move on flat ground should make a huge difference of 1 K, but given that other stations show a similar drop in 1950, it appears someone assumed wrong for De Bilt.
It takes a miracle that so many station simultaneously by freak error ends up with the very same freak “wrong” data graph. I hope all can see this.
On top of this, KNMI seem to hide raw data, and this obviously do not make them look good or trustworthy in any way. At least thats my opinion.
The Royal Netherland MI should openly share raw data from many many temperature stations with the Royal Netherland population.
Hiding data like CRU gives bad karma…
cRR Kampen,
If you have Raw Zwanenberg or any other dataset (from Europe or beyond) mail me on fel@nnit.com, thank you.
– especially If you have 30 years in the Met world, please send me raw data also from Belgium (That is EVEN MORE RESTRICTED when it comes to temperature data before 1951 than KNMI !!) please let me see it.
What is it w/Mosher on the lat/altitude stuff? If you’re working w/anomalies at single, unchanging stations, you don’t need to worry about it — anomalies are anomalies.
If you’re talking about single stations changing location, then yes. But then why not treat changing station locations as separate, unique stations?
Village I.
You write: ” A completely new, true, uncontaminated data set for Global Temperatures must be produced ”
Yes, you cant examine temperature before knowing which data are sick and which are good.
Its that simple.
I am in the process of starting up such a temperature index “RUTI”, and even though im not there yet, RUTI is build on open dicisions and data sources, that is, RUTI is in fact the only nearly reproducable temperature set that exists at this moment.
RUTI:
http://hidethedecline.eu/pages/ruti.php
K.R. Frank
Possibly off-thread, but vaguely linked to this curiosity, possibly via “coastal” versus “inland” issues.
Benbecula, out in the Hebrides, is a “land” site but I imagine is totally dominated by the Atlantic airstreams. Its BEST temperature record is here: http://i44.tinypic.com/2ues32p.jpg (The polynomial fit is for cosmetic and entertainment value and is not intended to have any predictive or analytical value.)
Over the last 160 years the air temperature shows a sort of cyclic change with an amplitude of half a kelvin and period of eighty years but I don’t know why. Any ideas?
Benbecula: http://maps.google.co.uk/maps?client=safari&rls=en&oe=UTF-8&redir_esc=&q=benbecula+map&um=1&ie=UTF-8&hq=&hnear=0x488d1df940b0184f:0xf0c681afa672330,Benbecula&gl=uk&ei=fVu2T5mdG8bN8QO2tuinCg&sa=X&oi=geocode_result&ct=image&resnum=1&ved=0CA8Q8gEwAA
The article already write: “Gemert had a large break in October 1949, when the station was renovated.” In the meta data on De Bilt, I found that there was a change from a Pagode Screen to a Stevenson screen in Mai 1950. The renovation in Gemert may have been a similar update. This screen change may well explain the artificial drop in temperature in 1950, which needs to be corrected to be able to compute a reliable temperature trend.
To measure air temperature a thermometer needs a screen to protect it against radiation and wetting. The Pagode Screen is open at the bottom, such a screen is shown in the meta data of De Bilt (this page is in Dutch, the Pagode screen is shown in the top two pictures at the bottom of the page):
http://www.knmi.nl/klimatologie/metadata/debilt.html
At the time they probably thought that an open bottom was good for ventilation. In other countries similar screens with an open bottom were used as well, such as the Wild Screen in many Northern and Eastern European countries and the Montsouri Screen in Southern Europe.
http://www2.meteo.uni-bonn.de/mitarbeiter/venema/themes/homogenisation/HOME/#photos
On sunny days with little wind, the land surface heats up, the heat radiation heats the thermometer (radiation error), which biases the measurement. Therefore Stevenson Screens (Cotton Region shelters) were used more and more, which are closed to all sides and use double Louvre walls. The difference between these two screens can be studied by making several years of measurements with both screens at the same location (parallel measurements).
Silver Ralph aks: “Give me one logical reason why all historic temperatures need cooling, rather than warming?”
Many inhomogeneities are random in sign. But the transition of screens with an open bottom to Stevenson screens occurred world-wide. If you go back even further in time, to the 19th century, temperature was typically measured at the north side of a building, near a window of an unheated room. Even if the room was unheated, this still also created a warm bias.
As these transitions in how temperature was measured occurred world-wide, they are seen in all the raw data and homogenization of the data will correct the too warm values to the cooler values that would have been measured, had people used Stevenson screens at the time.
By the way, in Eelde they had a Stevenson screen from the start.
http://www.knmi.nl/klimatologie/metadata/eelde.html
This may be the reason that the KNMI is willing to give out this raw data. As this series is relatively homogeneous, it cannot be so easily abused by climate sceptics to mislead uninformed citizens.
tonyb says:
I wonder if the constant fiddling with figures has its genesis in such studies of Historic temperatures carried out by such as Phil Jones. In a book of his I recently read, in which he re-analyses a number of the very earliest instrumental temperature records (18th Century) he made a comment something along the lines of;
‘The instrumental records we examined seemed to be showing warmer temperatures than our computer models indicate should have ocurred. We have therefore adjusted the instrumental record.’
Which book was that? And is there any chance you can remember the exact quote?
Here is a long term composite graph using, averaged anomalies (1960-1990 base period) including the Debilt, Central England, Upsalla & Berlin-Tempelhof data. These were selected, because they had records starting prior to 1750.
http://dc254.4shared.com/img/l2jEHPxe/s3/Ave4_20yr_3fil.gif
Three filter methods were used, with a 20 yr. ( 0.05 cyc/yr) “cut off”.
MOV- Moving Ave.
FR CH – Forward Reverse “filtfilt” (MATLAB) recursive 2-pole Chebushev
FF – Fourier Convolution
What is interesting, is that there seemed to little warming, in Europe in the beginning of the industrial age, and it was only after 1960, that the temperature “somehow” started to rise.
Here are a couple more. Ave1 is the CEL data using a Fourier 0.05 cyc/yr filter, while Ave14 uses stations that started recording prior to 1800.
http://dc457.4shared.com/img/k3wgtEsA/s3/Ave1_2010_FF_20yr.jpg
http://dc273.4shared.com/img/I04JY2jI/s3/Ave14_2010_FF_20yr.jpg
What was interesting about the Ave14, was that it showed the 1850 era was a local low point in temperature, & by coincidence, the starting point for the HadCRUT data sets.
PLEASE GIVE ME A TIP IF YOU HAVE KNOWLEDGE ON
Henry@frank
(are you Dutch?)
the problem is that if you go and look before 1970 you are running into a problem.
After 1970 most weatherstations were put on continuous temp. monitoring devices,
but before that they had to have people doing the readings of the temperatures every so often.
They usually did have have thermometers that measured the minima and maxima every day so you might find something usefull there if you go before 1970
but average temps. before is a definite no-no – everything become questionable and open to doubt.
People doing the reading of temps. is definitely a very an unreliable source.
Personally I found maximum temps giving me some good information:
http://www.letterdash.com/henryp/global-cooling-is-here
TonyG
If you go to this link you can find a developing conversation about this and the book name (I think).
http://judithcurry.com/2012/05/17/cmip5-decadal-hindcasts/#comment-201203
Coincidentally, as I saw your comment I had been looking in my digital files to see if I could find the exact reference as I think I emailed someone about it at the time (a month or two ago) as it was such an astonishing comment. If I come across it I’ll let you know
tonyb