Skip to content

Analysis Of UHI In South Korea

July 9, 2013

By Paul Homewood




I ran a post earlier about UHI in South Korea. A study by two Korean scientists found that over half of the warming reported in Korea since 1954 was a result of UHI. In many cities, the effect of UHI was to add more than a degree to temperatures over the period they had analysed, 1954-2008.

I was already aware that GISS tended to allow only about 0.2C for UHI in other cities I had examined previously, so I was curious to find out what allowance they were making in Korea.


There are ten stations currently in use in South Korea, which are in the GHCN database, that in turn is used by GISS. These are all listed in Appendix A.

First, a brief explanatory note. GHCN collect temperature data, which they then put through a homogenisation algorithm, which may result in adjustments. These adjusted numbers are then used by GISS, who then put then through their own homogenisation process, which is designed to make allowance for UHI effects – more detail here.

The figures below compare the raw data with the “after GISS adjustment” data, so the resulting differences are due to a combination of both GHCN and GISS adjustments.

The study only listed Pohang, Seoul and Mokpo in the abstract. The other cities, marked as n/a, may be in the full paper, but it is paywalled!


The following points are noteworthy:-

  • Out of the ten stations, only one, Ullungdo, is rural.
  • The typical adjustment from raw to “UHI adjusted” is much less than the study found.
  • The average adjustment for the nine urban sites is a paltry 0.05C
  • There is, incredibly, no adjustment at all for Seoul, while the adjustments at  Mokpo and Cheju have actually added to the warming trend.
  • At the only rural station, Ullungdo, the GHCN adjustment has reduced the 1954 temperature, thus adding to the warming trend.
  • It is often claimed that station moves out of city centres and into airports can offset UHI. This is not the case in South Korea, as only one site, Chunchon, is listed as an airport site on the GISS station list.



Since the 1940’s, there has been a massive increase in urbanisation in South Korea. The country’s population has risen from 20 million in 1949, to over 50 million today. Between 1945 and 1985, it has been estimated that the urban population increased from 14% to 65% of the total population, which would imply more than a tenfold increase in actual numbers.

On top of that, industrialisation and technology have made these cities unrecognisable from 50 years ago. It is inconceivable that these factors would not have created a significant increase in UHI effect over the years.

Which all raises the question, why are GISS allowing for so little? With only one rural station, their temperature calculations for the whole country are heavily skewed towards these urban sites, and therefore must be viewed with considerable scepticism.

What is true for South Korea is also true for much of the region, as mass urbanisation and industrialisation have similarly affected many other countries there, such as China. Has correct allowance been made for UHI in these?

Meanwhile Richard Muller tells us


Urban areas are heavily overrepresented in the siting of temperature stations: less than 1% of the globe is urban but 27% of the Global Historical Climatology Network Monthly (GHCN-M) stations are located in cities with a population greater than 50,000.







Location Population
x 1000
Temp Change
1954 to 2008
Temp Change
1954 to 2008
After GISS Adj
To Study
Pohang 134 0.63 0.33 0.30 1.17
Seoul 6890 1.26 1.26 0 1.16
Mokpo 193 0.54 1.13 -0.59 <0.20
Chunchon AP    * 141 0.71 0.58 0.13 n/a
Kangnung 85 1.76 1.60 0.16 n/a
Ullungdo <10 0.98 1.08 -0.10 n/a
Taejon         * 506 1.88 1.87 0.01 n/a
Pusan 2453 0.81 0.52 0.29 n/a
Yosu 130 0.54 0.23 0.31 n/a
Cheju 135 1.01 1.17 -0.16 n/a


* Records for Chunchon and Taejon begin in 1967 and 1969 respectively




1) Raw temperature data

2) GISS adjusted data

3) UHI Study by Kim & Kim

  1. July 12, 2013 11:55 pm

    The paper provides values for two more of these stations, 0.87 for Kangnung and 0.48 for Yosu (transliterated as Gangneung and Yoesu respectively, but the coordinates match the locations shown on the map in the paper, so it is highly likely that these are the same stations). The remaining values in the paper are 0.65 Busan, 0.16 Mokpo, 1.35 Daegu, 0.60 Jeonju, 0.69 Gwangju, 1.02 Ulsan, 0.00 Chupungnyeong, and1.10 Incheon. I know nothing regarding transliteration of Korean,but the differences from the transliterations used by GHCN seem too great to suggest any other matches.

    One further small point to note is that GISTEMP treats Ullungdo as urban, not rural. The GISS station list still includes the R/U classification used earlier, but the rural/urban classification is now based on the nightlights value, 12 in this case, at the end of the line, indicating a periurban location, processed by GISS as urban. The WMO coordinates for this station, (37.481, 130.899) fall close in Google Earth to a linked photo of the Met Office, situated on the outskirts of a small town. Using the preferred nightlights image instead of that used by GISS, which I believe is now described as deprecated by the source, a value of 17 is found, which brings this location further above the rural/periurban threshold.

    I’ll try to have a look at the GISS Korean adjustments in more detail, and I’ll post a further reply here if I see anything of particular interest.

  2. July 16, 2013 9:20 pm

    Quick test to see if I can post tabular data in a further comment. Can be deleted.
    d e

  3. July 17, 2013 2:00 am

    I’ve had a look at the GISTEMP adjustment for Seoul, which replaces the trend for the urban station, Seoul, with an average trend of overlapping rural stations within 500 km. These rural stations come from South Korea, North Korea, and China. The values shown at the start of each line below are the trends found (degrees C/century) for the last 55 years, to 2012, using the GHCN qca file dated July 09 2013. This trend shows sensitivity to the particular rural stations used for GISTEMP adjustment, as discussed further below.

    2.55 GHCN qcu (GHCN raw)
    2.51 GHCN qca (GHCN adjusted, GISTEMP input)

    3.13 GISTEMP adjusted. No exclusion (GHCN coordinates/nightlights)
    2.88 GISTEMP adjusted. No exclusion (WMO coordinates/preferred nightlights)

    2.10 GISTEMP adjusted. South Korea only (GHCN coordinates/nightlights)
    2.59 GISTEMP adjusted. South Korea only (WMO coordinates/preferred nightlights)

    2.78 GISTEMP adjusted. No North Korea (GHCN coordinates/nightlights)
    2.04 GISTEMP adjusted. No North Korea (WMO coordinates/preferred nightlights)

    2.95 GISTEMP adjusted. No China (GHCN coordinates/nightlights)
    2.83 GISTEMP adjusted. No China (WMO coordinates/preferred nightlights)

    Sensitivity testing by exclusion of North Korean and/or Chinese stations above is self explanatory, but the pairs of values, GHCN coordinates/nightlights and WMO coordinates/preferred nightlights, requires further explanation.

    Although Hansen et al 2010 claims that “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 location data, as used by GISS, includes substantial errors for a large number of stations, leading to incorrect classification of these stations as urban or rural when the nightlight radiance at a point distant from the station is examined. The impact of a rural station incorrectly identified as urban should not be serious, as the trend at this rural station, when treated as urban, should be replaced by the average of other nearby rural stations. Incorrect classification of an urban station as rural however is potentially more serious. If this urban station has a UHI component, this UHI component then becomes available to contaminate the trend at any urban station where this false rural station is used for adjustment. The second value of each pair uses the WMO higher precision coordinates instead of the GHCN coordinates for WMO stations. These may still contain errors, but most of the South Korean station locations have been recently revised, as have some of the Chinese stations.

    Furthermore, the nightlights image used by GISS is not that preferred by the originators, and I understand is in fact regarded as deprecated. Comparison of radiance contours for the two images at selected locations shows odd contour artifacts and location shifts for the image used by GISS.

    The South Korean rural stations used by GISS are Ulchin, Taegwallyong, Samch’ok, Songsanp’o, Haenam, Chomch’on, Yongdok, Wando, Inje and Poun. Using WMO coordinates Ulchin, Taegwallyong, Chomch’on and Poun are reclassified as urban, but Chupungnyong and Kanghwa are reclassified as rural.

    The North Korean rural stations used are Chunggang, Haeju, Wonsan, Jsohin (Zvosin), Hyesan and Kimchaek.

    The Chinese rural stations used by GISS are Chengshantou, Dandong, Tonghua, Linjiang, Caohekou, Ji’an and Changbai. Using WMO coordinates Dandong and Tonghua are reclassified as urban.

    I have listed these stations in descending order of record length for each country. GISTEMP selects rural station records in this order. All of the South Korean rural station records however are quite short, with only Chupungnyong starting before 1972 and none continuing after 1993. As a result, Chinese and North Korean stations, when available, are the first selected for this average rural record.

Comments are closed.

%d bloggers like this: