Skip to content

Urban Heat Island Effect In Japan

August 16, 2013

By Paul Homewood



It is worth taking a closer look at a paper I briefly referenced to earlier today, regarding UHI in Japan.



Urban warming in Japanese cities and its relation to climate change monitoring

Fumiaki Fujibe 

April 2010



This article briefly reviews urban warming studies in Japan, where many of the stations established by the beginning of the 20th century are located in cities that have undergone rapid industrialization. The recorded rate of temperature increase is a few degrees per century in large cities and tends to be larger at night than during the daytime. In some cities, the increase in annual extreme minimum temperature exceeds 10 °C century−1. On the other hand, recent numerical studies have revealed widespread urban warming around Tokyo and other megacities during afternoons of the warm season as a result of extensive urbanization that enhances daytime surface heating. An analysis using data from the dense Automated Meteorological Data Acquisition System network has shown that an urban bias in recent temperature trends is detectable not only in densely inhabited areas but also at slightly urbanized sites with 100–300 people km2, indicating the need for careful assessment of the background climate change. There is also some evidence of microscale effects on observed temperature, as revealed by an analysis of the relationship between trends in temperature and wind speed.


So the key points are:-


  • The UHI effect could be “a few degrees per century”.
  • UHI effect larger at night time.
  • UHI also widespread during afternoons of the warm season.
  • UHI is also detectable at “slightly urbanised sites”.


Let’s put into perspective the scale of 100 people/sq km.


Wikipedia give us a list of the top European cities by population density.




So, in cities we are talking in tens of thousands per sq km, even relatively small ones. Interestingly, the top of the list only has a population of 63000. (Notalotofpeopleknewthat!!)

A figure of 100/sq km, I would suggest, would cover most settlements, right down to small towns and villages.  Take for example the Japanese town of Haboro, that I used as an example earlier. It has a population of 8000, and judging from the station metadata below has an area of perhaps about 6 sq km.




That would yield a population density of at least 1000/sq km, for a small town that would normally be regarded as an rural site. I suspect that most towns of more 1000 would have a density of at least 100/sq km.

If the authors of this paper are right, it is clear that the UHI effect may be more significant than we sometimes imagine, particularly in countries where there has been rapid development in recent decades.

  1. August 16, 2013 9:33 pm

    Reblogged this on CraigM350.

  2. August 17, 2013 12:58 am

    Gistemp at least does recognise Haboro as an urban station, and did so even before introducing nightlights to classify stations. The station was rated “S”, or peri-urban.

    • August 17, 2013 9:09 am

      Yes, they have the population as 14000, perhaps including outer areas.

      However, they show no UHI adj since 1940, only a small one of 0.1C from the early years.

      • August 19, 2013 1:46 am

        However, if they were to use the preferred nightlight radiance image to determine rural/urban classification, rather than the deprecated one they currently use, 21047405001 Omu would then be classified as rural rather than urban, as it was before the use of nightlight radiance was introduced, and would be used as part of the rural record for adjustment, along with the other two Japanese and five Russian rural stations currently used by GISS.

        This results in an adjustment of 0.1C before 1970 and 0.2C before 1930. The trend (deg C/Century) is then reduced for the full record from 0.81 for the raw data to 0.52 for the adjusted data (0.60 as adjusted by GISS).

        The two nightlight radiance images may be compared at Omu nightlight radiance images. The preferred image is shown at the top, the deprecated image below. Heavy yellow contour is at value 10 (rural/urban boundary). Lighter yellow contours are at values 1 and 5 (rural). The coordinates for the station in the inventory file are just offshore, rural in the preferred image, peri-urban in the deprecated image.

      • August 19, 2013 6:05 pm

        It should not be necessary to point out that it is unlikely that this station is really located offshore, or that analysis including this station is “better” than the GISS one. The relevant lesson is that even a small error in location metadata can affect the resulting analysis for an individual urban station. (The many such errors in the station inventory, affecting the classification of stations as urban or rural, do in fact work their way through to affect the global temperature record derived by GISS)

  3. Brian H permalink
    August 17, 2013 3:13 am

    UHI contamination of records is far larger than anyone thinks.

Comments are closed.

%d bloggers like this: