134: Urban heat island (UHI) - implications for climate analysis

 

Fig. 134.1: UHI evidence from six cities in the Southern Hemisphere.

 

In my previous six posts I highlighted six examples of temperature records from major cities in the Southern Hemisphere where the temperature trend appears to be influenced by the urban heat island (UHI) effect that I discussed in Post 127. The result in each case was that the city exhibited significantly more warming than the rest of the state or country that the city was part of (see Fig. 134.1 above). However, the implications of this go beyond what is happening in a few megacities.

If large cities like Jakarta and Buenos Aires are warming because of urban heating, it then follows that smaller towns and cities must as well, only just not as much. As most weather stations are located near sites of human activity or habitation, it then also follows that most of the temperature records are likely to be corrupted to a certain degree by this UHI effect. The problem is determining to what degree that is. 

The six cities I have examined so far are obviously extreme examples and cannot on their own have a significant impact on their regional temperature trends as these regional trends are usually the result of averaging data from dozens of different stations. If they did have a large impact, then the difference between the trends for each city and its region would not be so pronounced. But that does not mean that the wider region is unaffected. 

The regional trends I compared each of the city trends with are also probably affected by the UHI effect, just not as much. That is because most of the data that goes to make up the regional trends comes from stations sited around other, smaller urban developments. So the regional warming is probably overstated as well, just not as much as that of the megacities. It is only because the megacities are so extreme that they stand out from their wider regions, but in reality both trends are likely to exhibit increased warming due to the UHI effect, just not as much as each other, hence the divergence in trends. 

And of course, if the temperature data for all the stations in the region is homogenized, then there is a risk of importing warming trends from the UHI into the surrounding area thereby raising the temperature trends of the wider region more markedly. Either way, the temperature trend for any region is likely to be corrupted to some degree by UHI effects in some of its stations.

But perhaps the biggest issue is the problem of area weighting. Ideally the contribution of each station to the overall regional average should be in proportion to area surrounding that station. So if a country has fifty stations and they are all evenly spaced, then each one should represent 2% of the area and each should make a 2% contribution to the mean temperature for the country. But suppose the station sits in a UHI and the area of the urban region is less than the nominal area that the station is supposed to represent. This is what happens with Jakarta in Indonesia (see Post 131) and for many other UHI stations.

The area of Jakarta is about 660 km², but that is only about 0.035% of the 1,900,000 km² area of Indonesia so the true weighting or contribution of Jakarta Observatorium (Berkeley Earth ID: 155660) to the mean temperature trend of Indonesia should be only 0.035%. But as there are less than sixty reliable temperature records for Indonesia (see Fig. 31.4 in Post 31) that means that each station on average makes a contribution of over 1.7% to the regional temperature trend. So a simple average of all station anomalies would over-represent Jakarta Observatorium (Berkeley Earth ID: 155660) by at least fifty times, and so too would an average based on a gridded method.

So how could we separate the UHI contribution to local warming from other climate drivers like carbon dioxide? One way to go about this might be to compare temperature trends for raw unadjusted temperature data from known urban and rural stations. But the problem here is that we don't know for certain where all these stations are exactly, or what the local geography around each station is. For example, according to its GPS coordinates Perth Regional Office is in the middle of the sea over 8 km off the coast of Perth. This is because the longitude and latitude coordinates for many stations are often only specified to the nearest 0.1° and sometimes the error is even greater. For Perth Regional Office the error is stated as ±0.2° for the longitude. That means its location accuracy is about ± 10 km. That is bigger than the extent of most urban areas. So finding sufficient pure rural stations against which to compare all the others is tough.

The takeaway here is that the urban heat island (UHI) effect is a big problem when interpreting temperature data, but determining how big is also a big problem.