This is an article by Roger Andrews.
A few years ago I did some work on urban heat islands and I’m taking the opportunity to document the results. [note 1]
Figure 1
One of the problems with urban heat islands is a lack of information on how large they are. Available data consist of temperature profiles across an idealized metro area, IR satellite images, maps showing temperatures on a summer afternoon or a winter morning or comparisons of temperature gradients at urban stations with surrounding rural stations. But there are no maps, or at least none that I could find, that plot absolute mean temperatures in and around a UHI. So I drew some.
Figure 1 shows the world’s largest UHI – Los Angeles, California, metro area population 12.9 million. (The temperatures are 1960-1991 means, which is the best I could do with the US records because a lot of them stop after 1991. But LA hasn’t grown all that much since then so the data should be reasonably close to present day conditions. Note also that temperatures in this and following maps are adjusted for zero correlation with latitude and elevation.)
Temperatures in the LA metro area average 17.9C and temperatures outside average 16.1C, a difference of 1.8C. Multiplied by the 2,300 square mile extent of the LA metro area this gives a UHI with a size of 4,140 degrees miles2 (or for purists 10,720 degrees kilometers2).
Second is New York City, Figure 2, metro area population 19.4 million. The New York UHI is concentrated around the downtown urban area, but if the entire metro area is considered we get a temperature difference of 1.0C (11.2C inside, 10.2C outside) over an area of 2,250 square miles, giving 2,250 degrees miles2. (Why is the Los Angeles UHI almost twice as large? Probably because LA is surrounded by mountains that confine heat within the LA basin.)
Figure 2
Third on the list comes Chicago, no figure, metro area population 9.8 million, with 1,200 degrees miles2, and after that Miami with 900 and so on and so forth.
What about urban areas outside the US? Well, there aren’t many where station density is sufficient to allow meaningful UHI maps to be drawn. Because of higher population density metro areas outside the US also tend to be smaller and the size of the UHIs correspondingly lower. An example is Osaka-Kobe-Kyoto, Figure 3, (metro area population 18.7 million) the UHI map for which is shown below (based on GHCN data). Mean temperatures inside and outside the metro area are 1.1C different, about the same as New York, but because the metro area is only a third the size (700 vs. 2,250 sq miles) the UHI is only about a third the size too.
Figure 3
At the other end of the list are a number of major cities that show no detectable UHI at all, including Atlanta, Cape Town and Manila. The best example I have is the Washington-Baltimore metro area (population 8.4 million), Figure 4 shown below. Apart from the 14.2C value in downtown Baltimore and arguably the 13.7C value in downtown Wahington it has no detectable UHI. Temperatures inside and outside the metro area are effectively the same.
Figure 4
Altogether I mapped UHIs around 43 metro areas, 28 in the US and 15 outside the US where there were enough data to work with (Ankara, Brisbane, Cape Town, Istanbul, Johannesburg, London, Manila, Melbourne, Montreal, Osaka-Kobe-Kyoto, Perth, San Juan Puerto Rico, Seoul, Sydney and Tokyo). I can’t discuss the results for each city individually, so what I’ve done to summarize them is plot temperature anomalies for all of the 900 stations in these 43 areas against distance from the urban center, Figure 5, with distances normalized to an “average metro area” which, if it existed, would be thirty miles in diameter and have a population of 6.6 million. There’s a lot of scatter, but averaging the individual points gives a respectably smooth curve except for the blip around 55 miles. This curve defines a UHI with an average amplitude in the metro area of about 0.5C and a maximum amplitude of maybe 1.0C that is detectable out to a distance of about 30 miles from the urban center, or about twice the radius of the urban area:
Figure 5
Now UHIs that have an average amplitude of 0.5C and which are confined on average within 30 miles of major population centers aren’t going to introduce much spurious warming into the surface air temperature record, and of course they will generate no spurious warming at all if the population center doesn’t grow in size over the period of observation. But here are some comparisons anyway.
First a plot of metro area population against temperature change between 1969 and 1999, Figure 6, calculated as the difference between 1964-73 and 1994-2003 means, using data from 382 GHCN v2 stations outside polar latitudes. If urban warming is a) significant and b) related to population we would expect to see a positive correlation.
Figure 6
But we don’t. Apart from the minor upward step between zero and 1.5 million population we see no relationship at all between population and temperature change (the red squares show average temperature change by population range).
Does a relationship appear when we segregate out urban warming gradients? A difficulty here is that most records from stations in urban areas either don’t show an urban warming gradient or the records aren’t good enough to say whether one is present or not. I have in fact been able to find only 22 urban stations in the world where an urban warming gradient can be reliably quantified by comparison with surrounding records, and the warming gradients at these 22 stations during the 20th century are plotted against metro area population Figure 7 below.
Figure 7
Again there’s no relationship (yes, that’s a dead flat trend line). There’s also no apparent commonality between the urban areas. Certainly you don’t often see Las Vegas, Moscow and Riyadh plotted on the same graph.
One would think that there would at least be a correlation between urban warming and UHI size, but the ten stations for which I have both UHI size and gradient data don’t even show that, Figure 8. The correlation is in fact negative. (Urban warming gradients are plotted against maximum UHI amplitude because the stations that show the urban warming gradients are usually at or close to the maximum UHI location).
Figure 8
So if there is spurious warming in the surface temperature record and population and UHI impacts don’t cause it, what does? The only thing that might is conditions in the vicinity of the station. To investigate this question I located the 900 stations used to construct the temperature vs. distance graph shown earlier as closely as I could on Google Earth and and segregated them, again as closely as I could, into three categories
- Urban – city centers, concrete canyons, heavy industry etc.
- Suburban – houses, schools, offices, roads, trees, gardens, parks, supermarkets etc.
- Rural – open country, maybe a few scattered buildings and roads.
and replotted the temperature vs. distance graph, Figure 9, with the data segregated into these categories. Here are the results:
Figure 9
Except for the dip in the suburban plot around 37 miles average temperatures at suburban and rural stations remain essentially the same regardless of distance from the urban center, with suburban temperatures running fairly consistently 0.2-0.3C higher than rural temperatures. Urban temperatures average almost a degree C higher than suburban temperatures and over a degree higher than rural temperatures in the zones of overlap. The implications as I see them are:
- Absolute temperature isn’t controlled by distance from an urban center but by conditions around the station. In other words, an urban heat island isn’t really an island at all. It’s an archipelago of urban and suburban heat islets.
- We don’t need an urban heat island to generate spurious warming. In fact we should probably discard “urban warming” altogether and use some more appropriate term.
- We can’t correct records for spurious warming using blanket corrections based on population or night light intensity. Each record has to be considered individually.
Note 1. A disclaimer, I can no longer find the link to the site from which I got all the US station data; you will just have to take my word for it that I didn’t make them up.
Tallbloke's Talkshop 
Roger Andrews says”you will just have to take my word for it that I didn’t make them up”
I Believe you! LoL 🙂
Looks to be way too much work to make all of this up. Besides Anthony Watts and others have done similar work by personally taking temperatures as they drive through developed / undeveloped areas. Your graphs look to be in line with their results. pg
Station heat islands?
Roger A, I go along with what you say.
I have an outside temperature gauge in my car. I have checked it in various places with displayed temperatures and with weather information of local radio, and also at home with other thermometers.
I live in a suburban-semi-rural area (properties around have 1 to 5 acres -most have lots of trees, good for bird watching). When I drive to the suburban centre (shops surrounded by 500-600 m2 house blocks with few trees) I find that the temperature is 1 to 1.5C higher night or day. When I drive to more rural areas (strawberry farms, orchards and larger grazing properties) the temperature is 0.5 to 1C lower -again day or night.
I would suggest that all temperatures are local and affected by the surroundings particularly the vegetation and tree density. I would think that any classification of sites has problems. Suburban Sydney (which I understand is the largest city in area in the world- certainly covers a bigger area than LA) with its surrounding and internal national parks (which are green all the year) will have a different UHI results to the mainly treeless Tokyo.
The other issue with the UHI effect I suggest is that most stations around the world even in rural locations have become more urbanised or at least are now in suburban surrounds over the last 80 years with huge increases in population (maybe 3 times greater) and the movement of people from rural to urban areas. This particularly applies to airports. In Australia, airports prior to 1960 (apart from the six capital cities and one or two in large centres) had dirt or grass runways at least 5 kms out from the suburban edge. Now they have bitumen or concrete runways and sealed parking tarmac around the terminals. Most of the airports have business centres for freight, hire cars, maintenance & manufacturing with housing now close by.
I have been gliding from one of these small airports. The glider could climb by circling through the warm air rising from the bitumen runway and could go long distances by flying over a nearby highway. I can recommend that as a great experience, and also learning about atmospheric conditions from the older experienced pilots. None the older pilots with long experience I met believe in CAGW but they do note the changes in terrain which they can use to their benefit. Note on a hot day one can actually see the heat haze over the bitumen runway. The experienced pilots can tell from the nature of the clouds and the movement of birds where there are good atmospheric conditions. It is exciting to be up at 10,000ft watching an eagle glide 20 ft away.
Purely anecdotal, but watching UK TV weather forecasts they often point out that night temperatures ‘in rural areas’ could or will be lower than shown on their summary charts.
Likewise the forecast London temperatures are more often than not a degree or two higher than the surrounding areas. What results might be had from a survey of rural stations only over a long period?
One of the problems in comparing data from “urban” and “rural” stations is defining which stations are really “rural” and which really “urban”. In most cases the metadata don’t locate the station accurately enough to allow us to do this, but still it’s not hard to identify “urban” stations that are miles outside the urban area they’re supposed to be in, such as Santiago, Chile, and Helwan (Cairo), and in cases where we can pinpoint the station location we find more than a few “rural” stations surrounded by concrete and corrugated iron, such as Tennant Creek, Australia (the station is at the yellow pin):
Even tougher is the problem of defining what the station used to be. Many stations with apparently continuous records have in fact been relocated multiple times, but records of the moves are usually incomplete at best.
Luckily, however, it doesn’t seem to make much overall difference whether the station is urban or rural. Figure 6 suggests that potentially spurious warming occurs about as frequently at rural stations as at urban stations, and indeed we see spurious warming in the “rural” record from Low Head, Tasmania:
http://www.john-daly.com/low-head.htm
But no spurious warming at all in the records from four “urban” stations in Europe:
If you want to get to a sensible model of urban heat island temperature changes, then there are several important factors to be considered.
1. Population density – clearly higher density will lead to more effect.
2. Energy use per capita – you can’t expect the same effects in India to those in USA.
3. Prevailing wind direction and speed.
When I look at the maps, the LA temperature increase seems to be a full 2 degrees, and NY not much less. I don’t know why you have so much lower figures.
Hi Ray:
Higher population density will have an impact on temperature if it results in more concrete and less open space, which it usually does. I’m not sure about mud huts, though.
Energy use ditto, although higher per capita energy use usually results in proportionately more open space, such as in the suburbs of developed-world cities.
Prevailing wind direction and speed? Dunno how to figure that into the equation.
None of these factors, however, are as influential as conditions around the station. Population density and per capita energy use aren’t going to make much difference if the station is in the middle of an asphalt parking lot, as the station at Phoenix Airport was for a number of years.
Another point that often gets forgotten is that an urban heat island by itself doth not an urban warming gradient make. The UHI has to get bigger and/or more intense over time to do this. This may explain why mature cities in Europe have UHI signatures but show no urban warming gradients.
My 1.8C estimate for the amplitude of the LA heat island (actually 1.87C now that I check) seems pretty close to your 2C eyeball estimate.
You can get ~2C from New York if you compare the concrete canyons of Manhattan (shown by the outline on the map) with stations way out in the boondocks, but in between are the suburbs and towns that make up most of the New York metro area. I should have outlined this area on the map too but for some reason didn’t do it.
Figure 8 shows a maximum UHI effect of 2.7C for LA and 2.2C for NY.
Roger thanks for your reply. Of course materials (concrete etc) are also important as you say.
Regards, Ray
Roger A: thank you for this interesting post.
You wrote: ‘One of the problems in comparing data from “urban” and “rural” stations is defining which stations are really “rural” and which really “urban”.’ and ‘In fact we should probably discard “urban warming” altogether and use some more appropriate term.’
Quite so, I couldn’t agree more.
If, (as seems to be the case) urban development influences local climate, then so does rural development. Everywhere where people live or work, urban or suburban or rural, they construct buildings as dwellings, workplaces and transport access points (airports, rail stations, ports). They build roads, railways, bridges. If such structures influence temperatures, humidity etc (and I’ve no doubt they have a measurable local effect) then the result, in reality, is a changed climate, which to varying degrees we are all subject to. The idea that one should somehow try to subtract this unknown ‘built environment’ effect to arrive at a ‘real’ global climate seems to me absurd and (unless one is a proponent of CO2 hysteria/warmist fanatic) pointless. The result of such an attempted calculation will unavoidably be spurious. The true climate is what we actually experience.
Coldish:
It’s actually not all that hard to identify reliable surface temperature records. We simply compare adjacent records to see if they match, and if they do we have a reasonable assurance that they’re not seriously distorted by spurious local warming or cooling effects. Here are five from an area covering maybe 50,000 square miles in and around Nova Scotia – two classified as urban (St. Johns and Shearwater) and three as small-town (Sydney, Gander and Charlottetown). The records aren’t identical, but they match each other closely enough to give us confidence that they’re measuring real temperature trends:
Now let’s compare the mean of these records with the only rural record in the area – Sable Island. Sable Island shows about a degree more warming between the early 1920s and the early 1970s. Why? We’ll probably never know. But it raises questions as to the reliability of the Sable Island record, so we don’t use it, or at least not the segment before 1970.
“Classic” urban warming effects aren’t that common, but when present they’re not hard to weed out either:
A modest observation from my location right on the edge of a small city. (And by “right on the edge” I mean that the front of the house overlooks open fields, while to the rear is a bog-standard suburban housing estate).
With the right conditions at night – clear skies, light winds and damp ground – I can watch mist forming on the fields, only to dissipate almost immediately when it drifts across the first row of houses. On occasions it is actually possible to look out the front window and not be able to see the houses across the road (about 20m away), while the view out the back is as clear as a bell.
An indication perhaps that UHI kicks in almost immediately upon the atmosphere encountering human habitation or construction. A point, IIRC, that Tim was making when he did his UK surface stations survey.
Spied one of those miraculous layers just rising from a large green. I suspect this is delicate, needs still air, wet ground, even trees stop it.
So far as I know there is no excellent weather site in a large urban area. A problem for example is needing to keep a very low cold horizon.
A more novel concept which I am not ready to formulate in any detail is the widely misunderstood concept of thermal coupling. It is this which causes the temperature delay. Urban areas are full of highly conductive “teeth”.
Actual heat generation is very minor unless a plume of hot air hits instruments.
The origin of all this is from the exercise I carried out with Lunar modelling. It became clear the usual ideas about Stephan-Boltzmann are wrong, it is no more than a theoretic construct which may not be applied in the real world but has to form part of a connection to the real world.
In that case I let it act upon a thermal impedance into a thermal capacity. That is all very well but omits a critical factor which is that for the problems in question it is never steady state, we are dealing with dynamic bodies under continuous change. Contrast with the static world beloved of climatic physics.
In the case of the moon the surface is covered with a fine dust under a hard vacuum and that means it will be a superb thermal insulator. This is what SB has to operate against. The moon also spins very slowly, everything slow motion.
Empirically this can be matched again measurements and what has been deduced about surface properties. All fits.
I’ve written barely anything about the situation for earth yet obviously I have looked.
The earth rotates fast. The situation is so complex and given the history of how it is impossible to keep people off pet topics I decided saying much was pointless.
One suspicion is a fit with Willis’ ~30C limitation for temperature, a figure which appears if there is a very strong thermal coupling, which makes sense given the earth properties.
Day and night cannot be ignored yet is widely. Seems to be a disconnect, only part of the whole gets considered. Albedo is a case in point… what does that say about nightside? Nothing.
Talk about clouds, mist, turbidity and yet there is poor data at best for nightside. Dark, can’t see, not many people looking. Space instruments struggle. Oh well, science is settled.
If urban areas increase thermal coupling maybe they will warm. Won’t be by much. Temperature profile will be different, and is but this is largely invisible in common datasets, time resolution is too poor.
No I don’t have answers just stuff to ponder.
“So far as I know there is no excellent weather site in a large urban area.”
Maybe Kew Gardens?
Ah yes, a station I have done but not published. (could have saved you the trouble)
I’m undecided on the worth of this station.
The collar of vegetation and elevated horizon including buildings is obvious. Photos do show photographic shortening, distances are larger than they seem.
Observe closely. Lack of gubbins on screen, is not Met Office synoptic station or auto. Turns out this is an early logger of dubious parentage which was removed and something else done. Photos predate this for a station which has changed. (original station is some distance away, closer to the Thames)
Station moves are not clear. I suspect quite a few moves.
Google Street view recently went there with the foot camera. AWS has vanished. Database tells a tale. (can email info)
Screen is close to north side of a small enclosure for manual reading. (reach over fence)
I’m concerned here about the very close proximity to the picket fence.
The logger (documentation of Met Office loggers does not fill me with gladness) has a nice hot PV up in the air. Not clear which sensor is which.
Wind mast etc. is farce. (hope it wasn’t used)
Might even be the case this is not Met Office equipment.
Digging as deep as I could showed the surrounds vary as gardens do.
London could have a good site in a Royal park except that when I publish on a particular site there will be a gulp on what I have unearthed. Explains why it doesn’t happen.
Ha! You ask for an excellent weather SITE. I give you an excellent weather site and you criticize the equipment. Talk about moving the goalposts. 😉
Grin.
What I know of the site so you got most of it, sorry.
On a bend of the Thames. I see it as an unstable site. Reasonably open at the moment but still not clear enough to rank as good enough.
Probably at the moment the 10 to 30 metres the Met Office want but these are lower standards than WMO
What is not clear is the details of the moved flower beds, pruning, growth, the station itself.
Some of this can be deduced from aerial photos. Kew was originally somewhat different so this would be the reference for UHI. History matters.
Even some of the apparently most stable sites on detail looking turn out to be not so good.
There is a similarly open station in Bradford. The details tell a story, could be good but.
Nice photo (copyright) which shows this in the frost and the effect of a steel enclosure. Used to be chains. And various other changes. What the stone cross does,..
The old Kew record, which stops in 1980, tells an interesting story in its own right:
(The four stations outside London are Rothamstead, Southampton, Oxford and Cambridge).
All those stations are dubious, this is England. As usual it’s detail which poses the problems.
We have things such as the clear air act, switch from town gas.
Technology changes, including switching standards. (around 1970 seems to have been a popular time to change to then WMO standard)
London was a collection of minor villages and towns, surprisingly little has changed. OPen areas tend to be connected with the State.
There are two Kew sites.
Kew ceased 1980. Kew Gardens still exists, the Observatory is now a golf course.
Google Earth is useful, invoke the history function.
The 1945 images show the area was surround by probably 1930s housing so there is little change. The Thames, a controlled river will have changed over time as upstream management took increasing hold. The area is also close to the huge water reservoirs which are upwind.
Developments upwind included Heathrow and surrounding complex.
I doubt these do much, bit far away.
This is roughly the WGS84 for Kew Observatory
51.4682 -0.31555
The 1945 map shows this cannot be correct and would have been close to the buildings.
Dates like that mean sometime around this dates +- 1 year
Station start date 01-01-1857
Station end date 31-12-1980
This is near enough the Kew Gardens site
51.4819 -0.29435
Station start date 01-01-1910
Station end date Current
The database information gives clues on some of what might have happened. Bit large to post.
Notes imply Kew went through changes, 1947 might be major, later on observer problems and falling from providing data, perhaps the drop you see is part of this, dates look about right.
And so on.
Someone who has the time and patience able to visit the various document places might be able to uncover a lot.
The observatory Google images show grass marks beneath the golf course implying there are old formal gardens underneath.
“All those stations are dubious, this is England.”
You do your country unjust. The Rothamsted, Southampton, Oxford and Cambridge records are excellent. Look how well they fit together. No artificial distortions worth mentioning here:
If you want dubious records take a look at Indonesia, or for that matter the USA, which probably has more dubious records than the rest of the world put together.
Even NASA named Atlanta – Hotlanta
http://www.warwickhughes.com/climate/atlanta2.htm
http://www.warwickhughes.com/climate/uhi2.htm