Skip to content

Urban Heat Island Effects on U.S. Temperature Trends-Roy Spencer

February 13, 2021

By Paul Homewood


Roy Spencer has carried out a detailed analysis of how UHI may be affecting US temperature trends.

He concludes that at least half of the claimed increase of 0.26C/decade since 1973 is due to UHI:




In short, he has ignored the USHCN network of stations, which have been heavily adjusted for Time of Observation Bias (TOBS) and homogenised, but are also affected by UHI. Instead he concentrates on 311 stations with hourly data, which thus avoids the need for TOBS.

These two graphs tell the story:




Fig 2 plots the temperature trends for all USHCN stations against the 2020 population density. There is a marked correlation between the two, the higher the population density the high the temperature increase.

A similar pattern emerges with the 311 ISD stations, many of which have low population density as they are at out of town airports.

Extending the linear fit back to zero population suggests the real US temperature trend, excluding the spurious UHI element, is only 013C/decade.

  1. JimW permalink
    February 13, 2021 1:25 pm

    Which is about the same as UAH satellite record of increase per decade, yes?

  2. February 13, 2021 1:42 pm

    Brilliant. Problem solved.

  3. Terry Breverton permalink
    February 13, 2021 1:45 pm

    I remember years ago with the fall of Communism, many, many Russian weather stations out in the frozen sticks were simply neglected and never then appeared upon official stats, artificially increasing Russian temperatures – by some distance and thereby showing a global temperature rise Does anyone remember this or can source this info?

    • Mack permalink
      February 13, 2021 7:24 pm

      I’d put that down to incompetence rather than design. Out of all of the climate models created by the world’s most celebrated meteorological institutions, hasnt the Russian model been just about the most accurate? And I do believe that the Ruskis are a tad cynical when it comes to CO2 being the main driver of modern warming. Many years ago they certainly embarrassed our very own Sir David King at a climate shindig, on whom HMG were then relying upon for the direction of future policy, when they rubbished his ‘global warming’ theories as completely unscientific. Not used to being called out on his theories, rather than debate his fellow scientists, he stormed out with the British delegation in a fit of pique. Class!

    • February 14, 2021 9:46 pm

      According to, …I think it was Dr. Richard Keen, the Russians out in the northern sticks used to report colder temps than measured in order to get the government to increase their allotment of firewood. Apparently temps took a jump up when those stations became unmanned or were closed, and the firewood was no longer needed.

      • February 14, 2021 10:45 pm

        Here’s the video.

        I’m pretty sure it was him, but it’s worth listening to anyway. I’ll listen again, to see if it was he who said it.


      • February 15, 2021 2:07 am

        I just finished re-watching it, and that’s not in there. It might have been Will Happer, or Richard Lindzen. In any case, that’s still an excellent video. I’ll see if I can find the other. It may take a while.

  4. Broadlands permalink
    February 13, 2021 2:19 pm

    NOAA-NCDC has gone back through all of the US. Weather Bureau official temperature data for the US 48 states from at least 1921 up to 1940 and lowered them systematically and most seasonally. The mean lowering is about 0.8°F. This is the opposite of what one might expect if the UHI effect is to be addressed. The net effect of these adjustments was to make 2012 the warmest year on record instead of 1921. The 1921 temperature was dropped from 55.6°F to 53.8°F. 2012 is 55.28°F.

    • Phoenix44 permalink
      February 14, 2021 10:40 am

      Lowering them systematically without demonstrating a systematic problem is fradulent. As I’ve said many times, if errors are random then over many measurements from many places they will average out and no adjustments are therefore necessary. If they are not random but some places systematically recorded too hot you have to prove that they did, not just assume they did. I have seen no evidence for the latter.

  5. MrGrimNasty permalink
    February 13, 2021 4:25 pm

    Some may have seen TonyB’s item on the Cambridge temperature record.
    Valid record or not (def. the latter IMO!), my impression at the time was that the heat around Cambridge that day was very much caused by the infamous W.London UHI blob flowing that way. In the heat last summer I recorded my highest ever home temperature on the day the wind swept the London UHI blob over us after several days of soaking heat.

    Current forecast is a spring-like 16C or more next weekend, prepare yourself for the headlines (but going by how much the forecasts change at that range day to day at the moment, might not happen).

  6. February 13, 2021 4:55 pm

    Sad to read that Germany had a bit of trouble with the cold weather
    Fortunately Bill Gates is coming to the rescue

    Maybe his book will mention how a lot less people die now due to extreme climate events than happened in the past due to good communication.

    The there is how a extreme climate event in 1709 affected our economy possibly even more that the governments covid restrictions

  7. Jack Broughton permalink
    February 13, 2021 5:38 pm

    While UHI is obviously a significant real effect that is abused in “global temperature calculations”, the graphs do not statistically prove or disprove the UHI effect. The data are almost random and the correlation coefficients too low to have much significance. It would be almost equally valid to draw a line with a slope of -0.03 deg K/decade through the data. I’m afraid that the main conclusion would be that it is not possible to prove the UHI from the data to any real level of significance from these graphs.

    Purely on inspection, it seems that the average temperature rise rate is 2.5 +/- 1.5 deg K/C.

    What seems more significant is that the raw data shows a mean of about 1.5 deg K/ century while the adjusted data shows a mean of about 2.5 deg K/ century. Assuming that nature is recovering from the LIA at about 1.5 deg K/century (the average CET rise since 1870), the only AGW is AlGorythmic.

    • February 14, 2021 10:54 am

      Jack, there is a very simple solution…..create a dataset using ONLY countryside locations and then making the comparisson…..Anthony Watts proposed exactly that in the US. Seems someone “needs” the urban data in the mix.

      • Geoff Sherrington permalink
        February 15, 2021 10:40 am

        Plerase see my longer comment below. Geoff S

  8. C Lynch permalink
    February 13, 2021 6:05 pm

    Joe Shute, the eternal propagandist in the Daily Telegraph, today grudgingly concedes the coldness of the Winter this year. He then says it’s not really that cold compared with the Winters in the Little Ice Age!?!
    That wouldn’t be the Little Ice Age that you and all the CAGW supporting “experts” have been denying existed for the last 30 years or so would it Joe?

    • MrGrimNasty permalink
      February 13, 2021 7:44 pm

      It isn’t by any definition a cold winter, modern or LIA, just a winter with, so far, some chilly spells in Jan, and a short cold snap in Feb. It will probably still end up in the top 50% warmest unless the current outlook forecast is dramatically wrong.

      It’s just MSM hype of ordinary events, like the nonsense that none of the Thames has frozen since 1963 when London hasn’t even gone much below -5C this year – total codswallop.

      Scotland/far north may have had a more markedly cold winter, but I doubt it is very exceptional (one very cold night is freak weather conditions, not a gauge of the season as a whole).

  9. Shoki Kaneda permalink
    February 13, 2021 7:40 pm

    The solution is obvious. Keep the CO2, get rid of the cities.

  10. Tonyb permalink
    February 13, 2021 8:23 pm


    Speaking of uhi, you will remember there was a fair bit of discussion here in 2019 concerning the heat record at Cambridge botanic gardens.

    I was able to get there last August and wrote up an article published today which covers the all time instrumental Heat record for the UK set in July 2019. I query whether the siting of the screen and the urbanisation around the site may make the record void.

  11. NeilC permalink
    February 14, 2021 6:19 am

    Thanks for this very detailed research Tony. Just from google map view the site looks suspicious especially since the building of the Sainsbury’s Centre.

    • February 14, 2021 9:49 am

      That is a huge structure. The heat being thrown off it by the sun was such that to get to the cafe we had to scamper past that building rapidly (technical term)

  12. Phoenix44 permalink
    February 14, 2021 10:44 am

    I live in London and used to work a few miles out of town. Winter early morning temperatures were always 1 degree cooler there. Ponds were frozen out of town, not frozen in my local park. I don’t know about Tmax but Tmin is hugely influenced by UHI.

  13. February 14, 2021 10:52 am

    There is no need to use any gauge which is located in a city or near one and subject to possible contamination anywhere today in the developed world. The fact that such gauges are used and then a fiddle factor applied (who controls those who control the fiddle factor) stinks to high heaven. We will all in the UK be familiar with being so often regaled with the new “record” temperature set at Heathrow Airport. Someone thinks we are stupid or at least ignorant and trusting and that applies to the majority of the population. I think it was Anthony Watts in the US who also showed this and proposed a station subset to be used which would avoid any issue with UHI contamination. His recommendation fell on deaf ears. He also showed that countryside locations where being systematically removed from the network and replaced by data gridded out from urban locations and what a surprise, up went the temperature!

  14. February 14, 2021 8:51 pm

    There was no temperature trend in UK time series from 1910 up to 1980, but after it became the consensus, a positive trend has been recorded.

  15. Geoff Sherrington permalink
    February 15, 2021 10:39 am

    Australia has a climate temperature record going back to the 1860s. Since about 1900, there have been hundreds of digitised station records to choose for research on matters like UHI. Australia also has a great deal more open space than many countries, so it lends itself to examination of the “urban minus rural T” approach to see what UHI there was.
    In 2011 I separated out some 44 stations that any reasonable observer would label “rural”. I have been to many of them. At the time, I even called them “pristine”. These were to form a basis of rural stations to compare with another selected set of “urban”. It never got that far. There was too much noise in the data. Too many missing values. Too much repeat data where a month of temperatures were copied to fill in another month of missing obs. That sort of thing. I chose 1972 as a start year because conversion of deg F to deg C happened that year; and 2006 as the end year because my digital data source ended in March 2007. Both Tmax and Tmin were analysed. I started with daily data, infilled, then averaged it to annual.
    Here are the numbers, in easy Excel form, for any readers here to crunch. It is easy to do.

    The more isolated the station, the poorer the quality, it seemed.
    There was so much missing data that infilling gave different final results depending on who did the infilling, or whether the missing data were just ignored instead of invented. These numbers here are invented by inserting a rough mean of the data just before and after the missing data on a daily basis.

    If you extend the time scale from 1972-2006 to 1972-2020, the trends alter substantially, some from negative trends to positive, some the opposite, for the small number of updates that I did recently.

    The data are not stable enough to derive much of interest. As the spreadsheet analysis shows, you can get the temperature trends to correlate with all sorts of extraneous variables. The trends even have a pattern when plotted against the stations numbers from the World Meteorological Organization, when they should not.

    If anyone thinks that you can extract useful UHI relations from these numbers, go ahead. I dare to say that you cannot. I cannot comment on the method of relating UHI to population, because just about all of these stations have essentially zero population within reach. However, from years of other UHI work, I cannot support the use of populations of people, of night lights or anything else. There is simply not enough recorded aboservation about enough sites with adequate quality to start to generalise and draw inferences. And, at least in Australia, there is not enough quality in the rural temperature data. It was never designed for this type of use and it is simply unfit for purpose. Geoff S

    • Jack Broughton permalink
      February 15, 2021 3:45 pm

      I think that your comment about the purpose and use of the data is the nub of the issue. The data (Tmax and Tmin) have been used for years to characterise the highest and lowest temperatures seen over 24 hours, which are important to farming, transport and health; this was the main use rather than the average value.

      In-filling data is only permissible for approximations and needs to be declared clearly. The AlGorythmic fudging of data being used to adjust history is Orwellian “goodthink”. Climate data are naturally of low signal to noise ratio which allows charlatans to draw semi-scientific conclusions, which as Booker noted are the basis for all fear-campaigns.

Comments are closed.

%d bloggers like this: