Skip to content

Cooling The Past In Siberia

May 30, 2014

By Paul Homewood


Центральная часть города.jpg


Krasnojarsk is a city in the heart of Siberia. According to Wiki, it currently has a population 1,035,000, a dramatic increase since 1939, when it stood at just 186,000. It is also said to be one of Russia’s largest producers of aluminium.


Map of Russia - Krasnoyarsk Krai (2008-03).svg


During World War II, dozens of factories were evacuated from Ukraine and Western Russia to Krasnoyarsk and nearby towns, stimulating the industrial growth of the city. After the war additional large plants were constructed: the aluminium plant, the metallurgic plant, the plant of base metals and many others.

It is therefore just the sort of place that you would expect a significant UHI effect over the years. Unfortunately though, it seems GISS disagree. On the contrary, they say that the rapid expansion and industrialisation has led to a lowering of temperatures.



Original Temperatures


GISS Homogenised Temperatures

The GISS adjustment can be seen clearly on the graph below. Temperatures in the 1940’s have been adjusted DOWN by 0.4C. From 1953 onwards, this adjustment has been reduced in steps.

As a result, a fairly flat trend has morphed into a warming one.





There are just four other current stations within 600 km of Krasnojarsk, and the UHI adjustments, etc for these are shown below. (There are two other stations which stopped reporting in 1990, and are both urban).


Location Population Brightness Index UHI Adjustment
Minusinsk 71000 17 -0.43
Enisejsk 18000 12 +0.10
Bogucany 11000 12 -0.20
Bratsk 246000 0 0


Sites are only regarded as rural if they have a Brightness Index of 10 or under, so only Bratsk is regarded as rural. Nevertheless, two of the three urban sites have their historic temperatures adjusted DOWN, in the same way as Krasnojarsk.

Worse still, Bratsk is regarded as rural, despite having a population of 246,000, and having been heavily developed after the war, in the same way as Krasnojarsk.


GISS calculate their UHI adjustments by comparing temperature trends for urban sites with those of rural stations within 1000 km. Ronan Connolly describes it thus:


To construct a rural average for an urban station, they require several neighbouring rural stations whose records at least partially overlap with that of the urban station. “Neighbouring" is initially defined as being within 500km of the urban station, but if that does not include enough rural stations, this is increased to 1000km. The contribution each neighbour makes to the rural average decreases as the distance from the urban station increases. If there are not at least three neighbouring, rural, stations with a common period of at least 20 years with the urban station, then they are unable to adjust the urban station’s record, and the station is not included in their global temperature estimates.


So the nearest “rural” station to Krasnojarsk happens to be Bratsk, 567 km away. There are two other rural stations, Bor and Bajkit, 639 and 661 km away respectively, so there are three stations to satisfy the criteria. However, because Bratsk is nearest to Krasnajarsk, this station makes the most contribution to the calculation. And the temperature trend at Bratsk?




Unsurprisingly, with such a hefty influence from UHI, the trend at Bratsk is upwards. As a result the flat trend at Krasnojarsk is adjusted from a flat one to a rising one as well.


This leaves the question – why is Bratsk regarded as rural? The clue lies in the Metadata.

The location given puts the weather station in the middle of the Bratsk Reservoir, with the result that the Brightness Index is zero.

The RussiaTrek website suggests a possible reason for the error:

On July 15th, 1955, the construction of Bratsk hydro power station and the city of Bratsk was announced as an All-Union construction. On July 26th, 1961, Bratsk reservoir began to be filled with water, September 1st, 1961, is considered the date of the Bratsk Sea coming into being.

Due to forming Bratsk water reservoir, the settlement Bratsk was transferred to a new place. In 1955, it was united with some other settlements and reformed into the city of Bratsk on December 12th, 1955.

It is likely that the original location of the station was in the original settlement, now under water, and that this information has never been updated.








Does any of this matter? After all, it is only one station, and surely these sort of errors cancel each other out? Unfortunately it is not as simple as that.

1) Wrongly identifying rural/urban sites tends to have a one way effect. If an urban site is wrongly classified as rural, not only is the UHI effect retained, but this effect is also passed onto other nearby urban sites, meaning that they are not properly corrected.

2) If a rural site is classified as urban, its trend is adjusted, and may even be dropped from the global record entirely.

3) From a spatial point of view, as there is much more rural area than urban, it is far more likely that urban sites will be wrongly classified than the other way round.

4) Because of the small number of stations in Siberia, the incorrect records for Bratsk, Krasnojarsk and the other sites mentioned have a disproportionate influence on a large chunk of Russia, particularly as they may be extrapolated up to 1200 km distant to fill in gaps.

5) There is plenty of evidence that these are not isolated mistakes, but even if they are they should not just be ignored.


Maybe the real solution is to ignore all urban sites, and calculate global temperatures from only genuine rural sites, that have all been properly vetted.

I will be passing this information onto Reto Ruedy at GISS, so hopefully this particular problem might be resolved.

  1. Pete permalink
    May 30, 2014 3:03 pm

    I would agree that global temperatures calculations, so we can understand any long term changes, should only be made from genuine rural sites. Using urban sites is mistaken.

  2. May 31, 2014 4:37 pm

    GISS uses the GHCN latitude and longitude values, many of which are inaccurate. This may not be a serious problem for GHCN – but it is for Gistemp, where classifying a station as urban or rural by examination of night time illumination is unreliable, to say the least, if you examine a location distant from the true location. Bratsk is only classified as rural by looking at a far distant rural location. The GHCN coordinates for Bratsk are for a location 110 km distant from the coordinates given by the WMO, and even the WMO coordinates are likely to be a bit inaccurate here, as shown on the right in the following image:

    Reto Ruedy is aware that there are errors, including serious errors, in the GHCN coordinates (“I’m not surprised at all that there are serious mistakes in this inventory file. It has been traditionally treated with less than the proper care”), but told me that “Unfortunately, we don’t have the manpower to check out all entries of that file”. While Hansen et al. (2010) states “Station location in the meteorological data records is provided with a resolution of 0.01 degrees of latitude and longitude, corresponding to a distance of about 1 km. This resolution is useful for investigating urban effects on regional atmospheric temperature”, this claimed resolution is often unreal.

    Crosschecking against the WMO data for GHCN stations which are also WMO stations currently shows 34 differences of 100 km or more, 8 between 50 and 100 km, 130 between 10 and 50 km, and 164 between 5 and 10 km, all distances which may be sufficient to give an incorrect urban/rural classification. I’ll add a further comment shortly on the number of stations actually wrongly classified.

    • May 31, 2014 6:27 pm

      That’s right Peter.

      GISS have the coordinate for Bratsk as 57.3N, which is in the middle of nowhere. I assume this is a typo though, as NOAA metadata shows 56.28N, which is in the middle of the reservoir.

      I have emailed Reto. To be fair, he has always been prompt in the past in replying.

      • June 1, 2014 2:16 am

        NOAA metadata showing 56.28N? Where does this come from? ghcnm.tavg.v3.2.2.20140530.qca.inv, contained in ghcnm.tavg.latest.qca.tar.gz just downloaded from still shows 57.28N, as does the qcu equivalent.

        I can well believe two different versions of the metadata coexist – data quality and consistency seem to be elusive targets for climate data.

      • June 1, 2014 10:16 am

        NOAA keep their metadata here.

        I assumed a typo, as GISS’s full list gives 57.28N, surely too much of a coincidence? (Their station look up rounds to 1 dec place)

        Station code is 22230309

      • June 1, 2014 1:02 pm

        As I said, I can well believe two different versions of the metadata coexist. Thanks for this second source – I’ll now try to compare with the version distributed with the GHCNM data files, which, being dated yesterday, I might have assumed to be the most up to date version, but it seems not!

        Possibly your source of metadata is a location where corrections are accumulated before being integrated with the regular ghcnm files. Odd though that they are not reflected in the inventory files which show frequent updated file dates to match the data files.

        No coincidence that the latitude differs by exactly one degree. This type of error is common among the mislocations. The WMO metadata has also shown the same type of error, but less frequently, and I’ve notified them ov such errors a couple of times, and they have been corrected.

        I’ve run a Gistemp analysis with the metadata and urban/rural classification corrected, and dug out a copy of the 2009 November Gistemp data files, the second last run with the old population based urban/rural classification, which of course recognised Bratsk as urban rather than rural. I’ll put together a post on my own blog showing these graphs for comparison with yours, and post a link here as I cannot post inline images in a comment here.

      • June 1, 2014 6:32 pm

        Thanks Peter.

        I’ll do a post on it

      • June 3, 2014 8:51 am

        I’ve put up that supplementary post now at There is quite a bit of further information about the adjustment process, and the effect of dropping Bratsk as an adjusting station (and the adjustment of Bratsk itself as an urban station), although I still have to add some more to the post, and in particular some discussion.

  3. May 31, 2014 4:42 pm

    The img tag was stripped out. The image is at

  4. Billy Liar permalink
    May 31, 2014 8:39 pm

    Seems to me that the thermometer is most likely at the airport which is at 56.4°N 101.7°E (and air traffic aside) appears quite rural.


  1. Boehner does belly flop on climate change - Page 2 - US Message Board - Political Discussion Forum
  2. Cooling The Past In Siberia – some supplementary information | Peter O'Neill's Blog

Comments are closed.

%d bloggers like this: