Invisible Elephants

Guest post by Tony Brown investigating the longest temperature records.  Tony has put together a well referenced post containing an interactive graphic which allows clicking on a location to see the individual records. WordPress free version won’t allow it to function so check out the link.   I think you’ll find it interesting.


This graphic

Pic Snapped by Lucy Skywalker

contains some 50 Instrumental temperature records that precede the 1850 Hadley Global temperature information. (Just press on a red dot on the globe)  There will be a well referenced study behind it in due course to put this information into perspective. It will hopefully become an invaluable resource for all researchers of climate.

These records provide a wealth of historic climate data for much of the Northern Hemisphere during a significant portion of the Little Ice Age. The datasets not only chart the considerable temperature variations through the centuries but also reflect the growth of the places they are located in, as many of the locations have developed from small towns in the 17th Century to large cities today.

I am currently looking for any more long data sets so this facility can be enhanced. Uccle (Belgium), St Petersburg and Lima (Peru) will be added shortly. I am looking for Cadiz/San Fernando and any others not already mentioned here. The criteria are pre 1850,which can be slipped ten years if it covers an area not already well represented.

The comprehensive study will tell much more, but after sifting through vast amounts of information and corresponding with a wide range of sources I would like to make a few interim observations.

There appear to be a herd of very large elephants in the climate room that are apparently completely invisible. The first is one called ‘global temperatures to 1850 that has a cousin 1880.

These global datasets are astonishingly complex, based on very small numbers of stations which continually change, and appear to be mainly a record of the 0.2% of the globe that has become urbanised, rather than represent the 99.8% of the world that isn’t. Therefore these creatures are to be treated with the utmost caution.

The next elephant is one called UHI. Many people seem to spend a lot of time looking the other way when UHI rampages by, including the IPCC.

This was the IPCC take on the urban heat island effect (UHI) in 2001 which hasn’t materially changed in SPM 2007, page.5:

Clearly, the urban heat island effect is a real climate change in urban areas, but is not representative of larger areas. Extensive tests have shown that the urban heat island effects are no more than about 0.05°C up to 1990 (from 1900) in the global temperature records used in this chapter to depict climate change. Thus we have assumed an uncertainty of zero in global land-surface air temperature in 1900 due to urbanisation, linearly increasing to 0.06°C (two standard deviations 0.12°C) in 2000.

Real Climate also seems to be wearing their invisibility glasses and is as equally dismissive as the IPCC.


The Urban Heat Island Effect has been examined quite thoroughly and simply found to have a negligible effect on temperature trends. Real Climate has a detailed discussion of this here. What’s more, NASA GISS takes explicit steps in their analysis to remove any such spurious signal by normalizing urban station data trends to the surrounding rural stations. It is a real phenomenon, but it is one climate scientists are well aware of and have taken any required steps to remove its influence from the raw data.


“The evidence points to a warming of about 0.6-0.8°C over the past century and a neglible effect on this from the UHI. While some ‘contrarians’ appear determined not to accept this finding, the evidence they cite appears thin indeed compared with the published research.”

World population has surged from 1.5 billion in 1900 to 6 billion now, for the first time in human history over half the population now live in cities According to the UN, the number of urban dwellers is expected to increase from roughly 3.2 billion today to more than 4.9 billion by 2030. Total area extent of the Earth’s urban land is said to be from 0.27 to 3.52 million km2 .Whilst Urban areas may therefore represent a small fraction of the globe it represents an increasingly large percentage of the instrumental global dataset.

Uhi has a disproptionately noticeable effect on those many stations that may have started off in a field hundreds of years ago, but are now in a green space hemmed in by buildings –these are very well represented in the pre 1850 datasets assembled in the graphic. (This link gives statistical data on population for each country/town going back hundreds of years)

The UHI effect was noted as far back as Ancient Rome. At 1.5 million this was a huge city by the standards of the time From statements by Pliny the Elder and entreaties to Nero to ‘provide narrow streets with high houses to provide shade’ the Ancient Romans were fully aware of the UHI effect, designed their city accordingly, and there are many records of the great and the good leaving it for the cool of the surrounding country in the summer.

According to this report “Even normal Greeks and Roman bought snow and ice imported on donkey trains. Few could afford private ice houses. Most urban residents bought it at snow shops. In Rome deep pits were filled with snow and covered with straw. Water melted and ran through forming a bottom layer of ice that sold at a premium. Snow could be more expensive than wine.”

It seems we haven’t learnt too much over the last two thousand years. UHI needs to be taken much more into account in that large proportion of urbanised stations in the global record, as the current adjustment appears much too small. Whilst Uhi is undoubtedly real, after reading some 30 studies I feel its effects can be exaggerated (other than on still clear nights)

City design, wind, land use and the siting of a thermometer amongst many other factors will have an impact on the degree of UHI to be applied. However, there is surely a practical limit as to how much a thermometer can be affected by UHI-heat will tend to be dispersed over a wider area as the urbanisation grows, not necessarily become more concentrated. Consequently I think this graph and calculation that follows seems pretty close to reality, albeit that high latitude countries are likely to show greater daily and seasonal temperature variability than is noted here, so a degree could probably be taken off the figures. However the logarithmic curve seems to be a more sensible representation of the effect of urbanisation on temperature than some other studies.

As temperature stations have moved (many to to airports) or are in locations which bear no relation to their original untainted position, so clearly the historic allowance for uhi-virtually zero- must be questioned. In this regards it is very difficult to support the offical position of the IPCC and Real climate that UHI is to all intents neglible on global datasets, as that includes so many urbanised stations.

The next Elephantprobably the largest and sleekest is that called natural variability.’ It is represented by the following quotes.

1 IPCC FAQ 6.2 Page114 of TAR4.

‘All published reconstructions find that temperatures were warm during medieval times, cooled to low values in the 16th 17th 18th 19th centuries, then warmed rapidly after that.’

2)   The UK Met office-a prime contributor through the Hadley centre to the IPCC assessments, assert:

Extract “Before the twentieth century, when man-made greenhouse gas emissions really took off, there was an underlying stability to global climate. The temperature varied from year to year, or decade to decade, but stayed within a certain range and averaged out to an approximately steady level.”

There is a great deal of evidence for considerable natural variabilty throughout our history, instrumental, written and observed. Commencing a data set at the depths of the last gasp of the Little Ice age (Giss/HadleyCru from 1850/1880) is all very well, but that there should be astonishment that temperatures have risen since ignores the undeniable fact that the current warm era is just one of the summits in a never ending series of peaks and troughs.

This excellent graph demonstrates the surprising variability of the LIA and its frequent warm periods. To paraphrase Dr Michael Mann the LIA is an outdated concept.

A cut off point of 1880 (Giss) or 1850 (Hadley/Cru) disguises the natural variability, as can be seen in two of numerous examples that can be observed in the ‘ Little Ice age Thermometers’ graphics-amply backed up by observations made at the time.

Giss global sets the scene for the familiar hockey stick

Fig A

Which is broadly confirmed by Giss from 1880-Hohenpeissenberg Germany

Plot of temperature vs. time

However this data goes back to 1781 for the same station

Temperature Graph for Hohenpeissenberg, Germany

It demonstrates that the escalating figure from the 1880 set is matched by a similar rise a hundred years previously and there are a very clear series of rises and falls

The second example also highlights that a start date of 1880 misses out on significant periods of climatic variability, but one which the Hadley set from 1850 catches. They come from two Swedish cities close to each other

This from Stockholm;

This from Arrhenius’ home town of Uppsala dated 1722-2005

Temperature Graph for Uppsala, Sweden

There was great excitement in Stockholm recently concerning rising temperatures to unprecedented levels, but a glance at the Uppsala chart demonstrates a warmer period just prior to the commencement of the Stockholm records (Both cities have had considerable study confirming a very real uhi factor) Consequently the Met office and IPCC assertion cited above is contradicted by numerous pieces of evidence.


* There appears to be an urgent need to carry out a comprehensive and independent audit of the global temperature records to 1850/1880.

* There needs to be a reappraisal of the undoubted natural variability of the climate that puts our modern experience into its proper context.

*The real effect of uhi on urban stations needs to be acknowledged.

*Someone, somewhere, needs to point out that a rise in temperatures from the low point of the little ice age is entirely to be expected and doesn’t warrant mass panic in the capitals of the world.

Therefore it seems appropriate that when the climate circus pitches up shortly in Copenhagen the descendants of Hans Christian Anderson-whose book ‘The Emperors New clothes was published in that city in 1837- will have the opportunity to point out the whole herd of apparently invisible elephants that is being ignored by a ring master who isn’t wearing any clothes.

Please let me know about any historic datasets pre 1850 you come across and please pay return visits to to watch the material develop.

Tony b

21 thoughts on “Invisible Elephants

  1. Facinating work.

    Another ongoing project connected with long-term temperature series is being hosted by EM Smith at the website “musings of the chiefio”. This site is well worth exploring. Smith has GISTemp up and running and is investigating some of the quirks of the programme. Of particular interest are cross-latitudinal trends in the spread of thermometers across the globe and changes in the mean elevation of thermometers with time.

  2. Rob R

    Yes I know about EM Smiths work and have made a number of postings over there. He is supplying some of the information that will eventually live behind the interactive globe on my web site. I think when referring to Giss the word ‘quirks’ is a very kind one!


  3. oh fantastic Tony. I’ve done a reduced-size pic of your brilliant interactive world map because that would be nice to head up your article, it’s in your inbox by now. I think you’ve got the gist of what we need. I’ll help as and when, you know my situation!

  4. What a coincidence! Yesterday I created my own “longest T record” graphic, meant to finally skewer AGW theory after so many years of thinking too hard about debate points. A silver bullet it really is now since it can be backed up by more than one record in a FORMAT that anyone who is unfamiliar with data sites and graphing software can see for themselves. I’ve thus added a link to that temperature page.

    Image is copyright Nik Willmore 2009 but I give permission for anyone to use it, unaltered. I will be happy to alter it if problems are found. I clipped 1659 due to one season being missing. The data is from averaging four seasons each year, plotted in Excel. Original data is:

    The result is deceptively simply looking but has been chosen with a graphic artist’s attention to detail and impact.

    Here is the Central England Don’t Panic Chart:

  5. With regards to UHI, you should read this
    Int. J. Climatol. 23: 1889–1905 (2003)
    Published online in Wiley InterScience ( DOI: 10.1002/joc.971

    Click to access HinkelEA-IJOC-03.pdf

    “During winter (December 2001–March 2002),
    the urban area averaged 2.2 °C warmer than the hinterland. The strength of the UHI increased as the wind velocity decreased, reaching an average value of 3.2 °C under calm (<2 m s−1) conditions and maximum single-day magnitude of 6 °C. UHI magnitude generally increased with decreasing air temperature in winter, reflecting the input of anthropogenic
    heat to maintain interior building temperatures."

    And Barrow isn't exactly a big city!

    Plus whilst the Met office may say one tghing about UHI, it's television forecasters say something quite different
    "These are City centre temperatures, out in the countryside, temperatures may be up to 5 degree (C) lower"

  6. Kondealer

    I agree that UHI has a very big effect-Nasa cited cases of up to 12C-so it must be right!

    However these are under very specific circumstances. In real life situations UHI will have a considerable effect on ‘average’ temperatures over a year of a few degrees-as you say the Met office frequently cite 3 or 4 degrees difference.

    I think the logarithmic curve cited in my article is a sensible estimate of the real life average effects, albeit that Northern Hemisphere circumstances are different to those in a small Australian town. I think there is a practical limit to the uhi effect at its upper end of 3 or 4 C (other than poor siting) and similarly I think the uhi effect is underestimated at its lower end. Even a 0.2 or 0.3 C allowance for UHI would make a considerable difference to our perception of modern temperatures.

    I will read the Barrow study with interest-thanks.


  7. Tony B, I agree that the log model is probably quite a reasonable representation. However it may well underestimate UHI in areas (typically high latitude, like Barrow) where considerable energy needs to be used to keep the population from freezing.
    This may will account for at least some of the “Polar amplification” seen in the temperature records, particularly the increase in night minimum temperatures.

    Maybe 2-3 different log models are needed to estimate (and therefore correct for) UHI in diffrent climate zones?

  8. Kondealer

    Absolutely agree with you. There needs to be a more realistic uhi adjustment made to global temperature records which is related to the individual circumstances of the micro climate being measured in that global record.

    Somewhere windy and at high latitudes or where the thermometer is in a very large park, will need a different adjustment to somewhere at a different latitude where a thermometer is hemmed in by buildings.

    The provision of 3 or 4 log models for different climate zones would seem to me to be a very worthwhile project for someone.


  9. The red Australia UHI/population graph miht be a bit overdramatised, but at lest its existence is admitted. There are other papers such as
    which give Melbourne (pop about 3 million)about 4-5 deg C, with othrs mentioning up to 10 deg C in modelling.

    But my concern is more with (a) the lower end of the population data and (b) the fitting of a curve.

    (a) Population is a useful but not a good index. This is because the weather station an be many km from the population centre and often is with small country towns. Qualitatively, I would not expect such a steep intial rise in the curve. Second, like in all experiments, one has to try to separtae or quantify perturbing effects from the effect under investigation. In the last 40 years, for a number of rural stations I have examined with no expectation of finding UHI, there has been very little change in Tmax or Tmin for the coastal ones in my subset, but there have been linear trends of up to 2 deg per century for a number of rural inland sites. Until this effect is understood, it is difficult to see how UHI can be separated out of the data from larger towns.

    (b) it therefore follows that the “one curve fits all” approach fails. BTW, unless I have skimmed Torok incompletely, there is no compensation for height above sea level. In any case, it also prevents a one cirve fits all.

    Denials of UHI from other countries, to be believed, have to be replete with reasons why it happens in Australia . Or forget them.

    For Tony Brown, I would be sure that you have the temperature records from James Cook on the Endeavour and Charles Darwin on the Beagle. One or the other, maybe both, was quite detailed. The do not last for long, but they are a window to overlay on other data.

  10. I think a very important reason to quantify UHI acceptably well, is to correct the really old records like CET. I think this can only be done station by station, hence a need for focus on a few long old records.

    Tony, I’m not entirely happy with the logarithmic UHI correction. I’d love to see the original cloud of data through which this curve was drawn, assuming it exists. Or a study producing a cloud of data, to produce a spline curve, if such a study does not yet exist.

    To really get a handle on UHI, I think we need to see the UHI factor quantified in other ways too: besides wind factor and altitude, I’d like to see variance (the spread of anomalies), seasonality (I expect more winter skewing), anomaly w.r.t. local summer AND winter mean temps (continental will give more winter skew; mean temps affect likely UHI), anomaly w.r.t. mean wind speeds, population density (effects of conurbations, city-centre parks, suburbs) and the economic-development metrics of Michaels and McKittrick.

    I don’t have the resources to do such a study, but flagging up the points is, IMHO, already advancing our understanding. Or does a study already exist, at least in parts?

  11. Tonyb said:

    “I would say you could draw another line from 1810 onwards-”

    That looked like overkill, whereas its absence makes the viewer actually mentally process the smaller trends (like you did).

  12. Nik

    Ah a cunning plan I see 🙂

    Lucy and Geoff

    I don’t disagree with anything either of you say, it just struck me that -however imperfect- a logarithmic curve is not a bad representation. Clearly there is an upper limit to UHI and clearly there is a lower limit.

    Amount of wind, height,proximity of roads,shelter, trees, water etc etc all play a role. However I think there is some kernel there and the curve is more believable than many of the numerous (30 or so) UHI studies I have read. If a station is in the middle of a big park in a windy place UHI is going to be fairly limited, one situated surrounded by buildings in a non windy place is likely to see a quite different effect.

    In reality EVERY location is different and needs its own factor but as far as I’M aware that doesn’t happen and the result is a UHI amount applied by Giss that seems unrealistically small.


  13. Geoff

    I have now read the paper you cited. It is very good. Obviously the UK’s climate zone is not the same as Australia’s, so there are important differences regarding the amount of heat that builds up in the day, night time temperatures, humidity etc. OKE is a much cited expert on UHI and I do tend to think his figures are a little exaggerated.(I’m sure my opinion will worry him!)

    However it all points to two things

    1) Giss/IPCC don’t allow enough for the many data sets that have become urbanised over the years.
    2) It is very difficult to see a co2 warming signal-the overwheming factors are natural variability, a UHI margin
    and land use changes.This latter may particularly apply in ther rural areas you mention.


  14. Nicely done!

    FWIW, one of my “pet peeves” about the GHCN, USHCN, and by extension GIStemp products is that a site is allowed only ONE flag and only in the present time for the Urban / Rural state.

    So a place like Sacramento would be flagged as Urban, because it is so today. But when did it START being urban? And how much has it changed over what period of time?

    GIStemp (and anyone else attempting a UHI correction) must make assumptions about when to start applying any UHI correction. It knows nothing about the past history of urban / rural state as a function of time.

    Clearly the UHI time profile of London or Rome will be different from that of Tucson Arizona or Sacramento, California or Las Vegas, Nevada (that was literally a ‘cow town’ until developed after water was available from Lake Mead / Boulder Dam / Hoover Dam post WWII ). Yet all of them are treated the same in GIStemp.

    The Urban / Rural flag ought to be stored for each year of the record (and, IMHO, ought to be a ‘size proportional’ flag not just an “is / isn’t” flag… )

    And frankly, on that issue alone, I think the GIStemp “UHI adjustment” is invalidated and ought to be thrown out.

    FWIW, there is a similar issue with the “Air Station” flag. Berlin / Temple was THE key airport of the Berlin Airlift. It has records going back to when it was an open pre-aviation era military field. It is now decommissioned and being turned into shops. Right Now, the record says it is an Urban Airport. In a few years, it will say Urban, not an airport and you get to assume IT NEVER WAS AN AIRPORT. In the past it was rural, be we don’t care about the past… Soon the acres of tarmac and concrete runways will fade into the past and be forgotten as well…

    And because this is in the base data, it impacts ALL users of the data.

  15. 13.Tonyb said
    November 8, 2009 at 3:02 pm

    Re fitting exponential curve.

    Simply, the curve does not contemplate cases where there is a small population and a weather station some distance away. This would require a start to the curve which was flat on the X-axis for a while before the rise started. The graph shown is also counter-intuitative because it assigns the greatest rate of UHI increase to the smallest increase in population.

  16. One of Sydney Australia earliest meteorologists was Henry Chamberlain Russell (1836-1907) and was based at Observatory Hill.
    BOM started 1901 so it may not carry earlier data.
    Still searching if there is pre 1901 data online.

  17. After further searching of the Australian BOM website.
    Search for Sydney Observatory site 066062 and that has data back to 1858

  18. Geoff

    Again not disagreeing with you-I am merely pointing out the curve is a good idea but is probably flawed at either end. I do not believe (as OKE does) that a hamlet of 10 people will raise the temperature by .75C or that a large city (as other observers may claim) will always raise the temperature by say 10C. Clearly both figures are absurd, other than in very specific circumstances.

    However a site next to a road next to fifty houses which previously was in a rural field is likely to show an appreciable increase and this will go right up to a practical upper limit for a city.

    But everything is dependent on numerous local factors.

    The continual shifting of stations does not help. Even a move of a few hundred yards can affect a reading. When stations move out of a town to what is at the time a small airport that then becomes a big airport, the original purpose of the reading-to record a specific micro climate- becomes lost.


    Thanks for your help


  19. Tonyb

    Re: Understanding Australia weather station startups. I have reviewed and sorted 1,700 stations that have (mostly) daily maximum and minimum temperatures..

    There were 15 stations to the end of 1859, (with Aust station number shown first; Melbourne data adjusted for change in station name):

    9017 FREMANTLE 1852 WA
    86071 MELBOURNE 1855 VIC
    70037 GOULBURN 1857 NSW
    58063 CASINO AIRPORT 1858 NSW
    63004 BATHURST GAOL 1858 NSW
    88029 HEATHCOTE 1858 VIC

    The numbers added in each subsequent 5-year term were, rounded to the nearest 5:

    1860-1865, 20 stations
    1866-1870, 30
    1871-1875, 35
    1876-1880, 60
    1881-1885, 85
    1886-1890, 105
    1895-1900, 70

    1901-1920, 125 stations
    1921-1930, 60
    1931-1940, 75
    1941-1950, 115
    1951-1960, 100
    1961-1970, 225
    1971-1980, 120
    1981-1990, 135
    1991-2000, 240.

    Note that many of these are at comparable locations, often removal from a town post office to a town airport, especially following WWII.

    These 1700 stations are rather variable in their continuity and duration and pointed to the need for a high quality Reference Climate Station (RCS) network in the early 1990s.

    It is difficult to extract the closure of stations because of the likelihood of name changes as above. Thus, it is hard to give figures for average durations of station records of good quality.

    These data are assisted by a good knowledge of Australian history and geography, especially concerning the gold rush years commencing 1851 which greatly led to population spread and the funds to afford to keep weather records.

    Data from the Bureau of Meteorology, Australia, with recognition of the effort involved in compilation. Any transcription or counting errors are mine.

  20. Did my fingers really type “counter-intuitative”? Sorry.

    It is hard to contemplate a one-fits-all curve. There are so many variables not even mentioned. e.g. variation in the height above ground of the thermometer gives a fairly large bias. At airports it is not uncommon to mow the grass. This has a partial effect of altering the height of the thermometer above ground, since most of the wind blows above thick grass rather than through it. I mention this odd effect because it has the capacity to cause large errors, of the order of half a degree C I’d guess, between new-cut and needing a cut in lush areas. (There are some studies in the literature of thermometer heights). One can envisage myriad effects like this, which make it frustrating to consider if an accurate reconstruction of land surface temp will ever be possible. Alternatively, the error bars are likely to be much wider than currently depicted.

    Remember that pan evaporimeter readings improved after placing a wire mesh over the water so birds did not bathe and splash.

    Do have a look at the vertical temperature profiles in the thesis
    “Optical Turbulence On The
    Antarctic Plateau”
    Tony Travouillon,School of Physics
    University of New South Wales, September 2004

    We’ve read of problems of snow depth around (and above)thermometers, similar in a way to the grass effect above. But the point of referring you to this thesis is the rather large difference between surface temperatures at stations and air temperatures from sondes. It makes satellite reconciliations, lower troposphere versus surface, a somewhat dubious objective, because of katabatic winds and inversions.

    So even reliance

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