From Dr. Roy Spencer’s Global Warming Blog
by Roy W. Spencer, Ph. D.
This is the third in my (never-ending, it appears) series on measuring the effect of Urban Heat Islands (UHI) on land surface temperature trends.
In Part I and Part II I emphasized the Landsat-based “built-up” structure dataset as a proxy for urbanization, which I’m sure we will continue to examine as part of our Department of Energy grant to examine (mostly) satellite-based methods and datasets for testing climate models and their predictions of global warming.
Much of the original research on the UHI effect (e.g. T.R. Oke, 1973 and later) related warming to the total population of towns and cities. Since population datasets extend back in time much further than the satellite period, they can provide information on the UHI effect going back well before 1900. In the last few weeks I’ve taken a detour from using the Landsat-based diagnoses of human settlement built-up structures as a proxy for urbanization, to population density. Along the way I’ve had to investigate issues related to low correlations, and linear regression (specifically, regression dilution). I decided not to cover that here because it’s a little too technical.
The deeper I dig into this project, the more I learn.
Urbanization Effects from 1880 to 2015
I have a lot of results I could show, but I think I will introduce just one plot that should be of interest. Using tens (in the early years) to hundreds of thousands of 2-station pairs of temperature differences and PD differences, I sort those from the smallest to largest 2-station average PD. Then I perform regressions in separate PD intervals (12 to 19 of them) to get the change in temperature with population density (dT/dPD). These coefficients are, in effect, tangents to the non-linear function relating PD to the UHI warming effect. The data shown below are from the month of June in 20-year intervals from 1880 to 2015, in the latitude band 20N to 80N.
Then, by summing those regression coefficients up (integrating them, in calculus terms) from zero PD to the maximum 2-station average PD value, I construct curves of PD vs. UHI effect. I have looked at quite a few published UHI papers, and I cannot find a similar approach to the UHI problem.
I have to admit, the results in Fig. 1 are not what I expected. They show the total UHI effect being stronger in the late 19th Century, and weakening somewhat since then. (Remember, because these results are based upon 2-station differences, these are spatial relationships, that is, for the 1880, 1890, 1900 period there is a greater temperature difference between rural and heavily populated locations than in later decades.)
I do not have a ready explanation for this, and ideas are welcome.
If the results were reversed, I would guess it is due to larger errors in early population estimates, since errors in the independent variable (PD) reduces the regression slope (dT/dPD) below the “true” relationship (regression dilution). But just the opposite is happening. And, it cannot be due to much lower numbers of stations in the early periods because that leads to only noise in regression coefficients, not systematic bias.
Some Thoughts
From reading the literature, I think this is rather novel approach that avoids a common problem: the usual separation of stations into “rural” versus “urban” categories. Because the curves in Fig. 1 are non-linear, a nearly rural station will experience much more warming from a given increase in population than will very urban site. Thus, previous investigations that found little difference in temperature trends between urban and rural sites don’t really prove anything. My methodology avoids that problem by constructing curves that start at zero population density (truly rural conditions).
Eventually, all of this will lead to an estimation of how much of the land warming (say, since 1880) has been spurious due to the Urban Heat Island effect. As I have mentioned previously, I don’t believe it will be large. But it needs to be documented.
I have often wondered if flood control / storm sewer systems in urban areas are influencing temps based on soil moisture content. Example: Cities leader responding to voters complaining about wet basements approve improvements in the canals which dries out the land sooner. Albuquerque NM implemented a watering/lawn ban on residential properties in the mid 80’s or there about. First on all new construction then on current properties. Is there enough temperature change in the metro area to detect the enforcement of these ordinances over time?
Could the decreasing rural/urban difference with time be due to an increasing amount of
improper siting of weather stations in rural areas ? Station locations close to buildings, over concrete slabs, asphalt, or rock will systematically raise the nighttime temps as well as the daily average temps. In the Urban setting, a lot of stations are probably badly sited anyway.
That thought had occurred to me as well. That is most of the UHI at a specific measurement point comes from say a radius of a few hundred and development further away has less of an effect.
As for why the PD effect was more pronounced in the late 19th century than now, towns back then were built much denser to be close to the RR station and a given population is now more spread out now.
I think Roy is very much on the right track in separating local warming from global warming. Countering the effects of local warming calls for very different measures than global warming – i.e. CO2 is not a significant issue with local warming.
Seems there is a hidden time factor in the graph, unless I misinterpret the graph. For example for an area to grow from zero PD to 2000 staring in 2000 could happen in a year. I live near Dallas and towns at the limits of commuting spring to life amazingly rapidly. The population center provide the residents. For that to have happened in1800 would have been impossible.
So does time create a new variable? Possibly, does the data for modern growth that happen so rapidly suffer because temperature equilibrium is not achieved during rapid population growth, whereas the slow growth of previous times allowed for full temperature stability.
You are very correct. Dallas suburbs can explode in population VERY quickly. Little Elm went from 2300 population to 50,000 in just 20 years. Farmland to concrete in just 2 decades. Frisco had 35,000 population in 2000 and now has 210,000.
I believe that an accurate example of UHI would be these rapidly urbanized suburbs. A 20 year snapshot of the Little Elm and Frisco, TX would be an ideal data set. From farmland to urban in a well defined 20 year span.
There is another growth phenomena that did not exist years ago and that is the rapid growth of retirement communities.
If temperature growth does not match rapid population growth because the causes of the UHI development lags the population growth then there would many examples to evaluate.
BTW, I moved to Plano when there was only one restaurant, a hamburger stand, and a population of 9000 people. Today the population is over 300,000. I got to pay through my taxes for all the roads, schools municipal buildings whatever.
In my opinion, population growth would always lead UHI. Current road infrastructures in growing communities will support much higher populations and traffic. Once the population reaches a critical amount, new roads will then need to be constructed. I have yet to see any suburb that has road infrastructure that leads population growth. Commercial business growth follows the increased population as well. Rooftops bring businesses to serve the increased population.
I do not have a ready explanation for this
I salute honesty like that. One thing about the 19th century to note is the amount of horses etc for transport among other things. They disappeared from an urban setting with the advent of mechanised movement. I wonder how much urban stables etc added to the equation?
Hear, hear
The 1880-1900 era was horse and buggy days…so comparing to today is apples and oranges. The automobile usage and widespread electricity usage made a big difference in heat produced per person. Look around your home at all the heat producing items beginning with HVAC and your automobile. How much heat per person is produced on average?
From a long ago post – February 26, 2010 at 7:22 am
I could never understand how UHI was minimized. If you look at New York City as an example.
Area, including water 468.9 sq mi ( 2,590,000 sq m)
Power used (2008) 54,869 GW-hr
(http://www.nyc.gov/html/planyc2030/downloads/pdf/progress_2008_energy.pdf)
Watts/sq m = 2,416 total. The Mayor says 80 percent is used by buildings and therefore 100 percent ends up as heat loss. Clarifying the 80 percent used by buildings was for lighting and heat, so by next day at the same temperature it was all turned to waste heat.
So the forcing is 1,933 W/Sq M
The file also remarks that the city has seen a 23 percent increase in the last 10 years, which is close to the increase showing up in the charts.
Also this doesn’t include a major business of NYC of food prep or transportation. All the waste heat from autos.
So does the +++1,933 W/Sq M affect more that +150 ppm CO2?
Whenever a new cause of any change begins I would first expect a rapid initial build of any effect to a plateau and then a decline in the rate of change as further incremental changes have progressively less and less effect.
Is that what the figures suggest ?
Roy, this is just half-baked thought on the surprise result. Perhaps with rising population at various sites there is a greater broadening of the UHI effect than thought, which means a so-called rural site has a small but unrecognized component of UHI that is attributed to its climate warming. A half-baked analogy would be pouring two spaced-out viscous tar deposits over time which would eventually coalesce at the ‘rural’ margins.
Population centers end up with radiating roadways, farmsteads, … that weren’t there before. Maybe some insight might be gained by adding a term that would straighten out the curves for further analysis. Also, since the UHI effect is steeper with rises from small initial populations, maybe the few farmsteads that appear have a greater effect than expected? And just maybe, this comment is no help at all!
Roy, I think Gary Pearse is on the right track. Very rapidly expanding suburbs from small populations to cities would be an ideal subject of UHI study. We had Dallas suburbs (Little Elm, TX) that grew from 2300 to 50,000 in just 20 years. Rural to urbanized in just 2 decades. That would be an ideal UHI case study. Frisco, TX grew from 35,000 to 210,000 in that same 20 year period. These two cities should render great insight into how UHI functions in reality. One area goes rural to a city and the other a small city to large city in the same 20 year span all measured with modern equipment.
Two possibilities: (and, or either)
1/ As the population density rises, so does the (temperature) of the UHI
As Jozef Stefan tells, as the temp rises the rate of energy loss skyrockets so you would get that exact shape of graph – the temp will rise with increasing PD but every doubling of (absolute) temp will lead to energy loss via radiation increasing 16-fold
On the same tack, is there simply more asphalt on the ground.
Yes it is black, absorbs a lot of sun and gets very hot.
That being the point = a slab of black asphalt at 70°C will be radiating at nearly 800Watts per square metre – almost a mirror as far as El Sol sees.
Ain’t that crazy, the asphalt might actually have a cooling effect
2/ And on to my pet rave, esp that there is now a Rural Heat Ocean out there because of much greater extent of intensive farming.
Farmers have lowered the albedo and also dessicated the soil.
Also got rid of trees and hedges (=things that cool) and = things that slowed down and got in the way of the now monster sized machines they use.
Consider that damp soil has a Heat Capcity of about 1,600J per kg and dry soil only 800J/kg
So you get a double whammy heating effect – the soil behind a plough absorbs vastly more solar energy than a green field but also needs only half as much energy to raise its temp per unit of temperature.
The essay and comments say that – the population density means more people means more food means more ploughing AND = less horses.
Horses needed perennial pasture for grazing and hay-making but the land they occupied around the cities has been turned into annual grassland.
i.e. Land that is only green for 3 months at most per year compared to pony pasture that was green all the time.
And horses didn’t need 4, 5 and 6 lane motorways – where’s The Jury on that now?
Does asphalt cool, or warm or both?
It certainly creates epic high temperatures but while doing so effectivly takes on an albedo approaching unity.
Hence why I keep saying that (no-one else can seem to get their head around the idea) ‘Deserts are cold places‘
I’d assert that cities would be also without the input of as much (fossil fuelled) energy as they get to hold them up at night.
After dark, use Wundergound or your favourite weather website to visit some deserts to see what I mean.
Deserts are cold places,
up at KKMC at the Saudi -Iraq border back in the late 80`s wake up in the morning to rocks cracking like rifle shots as the suns rays hit , with frost on the ground.
right. only during periods of green growth is there transpiration and therefore the enhanced heat flux aloft. the perennial grass climates selected by the ecosystem for the prairie. Traditionally a deep sponge organic layer with endless moisture reserves.
Soils eroded back to sand is one thing, and soils eroded back to clay-loam another. the texture of the mineral deposits influencing the heat capacity of the desertified terrain.
Logarithmic relationships. The law of declining marginal returns. The effect of adding 1 unit to 1 unit is 50%, 1 unit to 2 units is 33%, 1 to 3 25% … 1 to 1000 0.01%.
Same thing is true of adding a ton of CO2 to the atmosphere. That is why the fundamental equation of global warming is stated as Equilibrium Climate Sensitivity per 2X CO2, and the Temp at concentration T2 is the temp at Concentration T1 plus the ECS mulitplied by the base 2 logarithm of the ratio between the concentrations at T2 divided by T1.
T2 = T1 + (ECS * (Log2 (CT2/CT1))
Graph Temperature vs Concentration. The trace, like all logarithmic equations, arcs over and flattens out.
I just finished reading through the 32 comments submitted thus far. Clearly your comment is the most sophisticated and most likely the ah-ha Dr. Spenser may find helpful, thinking “of course, slipped my mind.”
But the important takeaway from this is substantiating proof that surface-based temperature reporting is helpful in the morning when deciding whether or not to wear a sweater but nothing beyond that level, Certainly, it has no value for climate policy making with $trillions at stake. Surely, this is an example of “The courage to do nothing”, Ignore it we have satellites, give up on the 1890 technology,
SWAG: Most energy use is industrial. As time has gone on, people live further and further from industry. (They commute.) Therefore, the energy use and population density becomes slightly more disconnected over time.
One thought that comes to mind about what might be affecting the UHI difference over time is that perhaps as cities get bigger, their edges are not as sharp, so if you are using the same pair, the difference between the “urban” and the “rural” one get diffused.
Another possibility is that as travel gets easier, “rural” stations become less rural.
It doesn’t take a lot of development before UHI gets noticeable.
One question immediately comes to mind. Do you use the average of the low and high for the day, the high only, or the low only. I remember in previous posts, with supporting data, that the highs changed little, the lows much more, thereby raising the average mostly because of the change in the lows. I would be interested in seeing a graph based on highs and another based on lows, to see if one dominated the changes you graphed.
It might be interesting to investigate how rural life has varied over the 150 years.
I suspect rural homesteads have a lot more stuff today compared to then.
Bigger homes, more concrete. How does the switch from horses and mules pulling the plow to using tractors affect land use? I suspect that even if there isn’t more concrete, there will be more areas with gravel instead of grass.
Echoing and maybe expanding on other’s comments. I’m wondering if there is a difference in instrumentation, recording technolgy methodology, and immediate siting and location at or before 1900. I find it difficult to believe in standardization before 1900, and around then automated graph recording technology would have become available for some instruments, and electric lighting would have come into use. Methods and procedures would have changed drastically as technology changed and was distributed.
If there is an effect from industrial thermal energy sources that would be about the right time for changes to appear.
Dr. Roy, my thoughts, sorry no calcs…
1) 2000-2015 is possibly lower because of more careful placement of weather stations as new boxes are installed (even at old sites). The others lines are too close to each other to assess.
2) increase with PopDens due to building height in the vicinity of many weather stations increasing decade by decade, and blocking the %ge of the cold sky seen by ground level objects, thus allowing them to be warmer in general.
3) Increased building height causing more ground level wind turbulence thus defeating the formation of nighttime temperature inversion of cool ground level, warmer air above by morning.
“Eventually, all of this will lead to an estimation of how much of the land warming (say, since 1880) has been spurious due to the Urban Heat Island effect.”
Clear thinking is needed about this. The first task is to measure the land warming. As a matter of measurement, UHI isn’t spurious. It happened, and should not be separated before measurement. Later, if you want to separate a UHI component from, say, an AGW component, you can do it.
The proper complaint about UHI is that location of measurement points in towns might bias the measurement. But I don’t think that is the emphasis here.
UHI is spurious (i.e. not being what it purports to be) if/when land temperature records are used as evidence of an enhanced greenhouse effect due to human-emitted CO2. Measurements in towns or near any local artificial (i.e. not greenhouse effect) warming do bias the measurement, and that is exactly the emphasis here. If not, that what?
“ if/when land temperature records are used as evidence of an enhanced greenhouse effect”
Maybe. But you have to get the record first, including a proper global averaging. Then you can have those arguments.
Sounds like you are admitting that the ground based temperature network is not fit for purpose.
Not fit for purpose, precisely, how can PhD scientists not “get it”. Ignore it, use the satellite data.
A confounding factor for the early periods could be that in the US the weather stations followed the telegraph and the telegraph followed the railroads.
I recall on this blog a comment that the early weather monitoring was the responsibility of railroad station masters.
Just as weather stations at airports measure increase airport activity the early data may be measuring increased railroad activity.
I suggest the weakening of the Population Density/UHIE is a feature of Inverse Threshold Effects: the quantum impact of the first changes is greater than that of the later ones.
That could be because initial urbanization is centrally strong, the area of further urbanization is lessened as a function of Area = PiR2. Or just the initial heat sources and heat sinks are somehow (roads/sidewalks?) more areally dense than later (“domicile urban) ones.
Or Standard Threshold Effects: there is no response until a Threshold is reached, when Blam! it shows up, thereafter having a lesser effect.
A friction type thing for a miving car: until static friction is overcome, nothing happens, and then big movement happens, when kinetic friction kicks in and dominates, until aerodynamic or other Effects become significant.
Spitballin’ here.
When there are fewer people than everyone ultimately expected to cause all the environmental chaos that was over-predicted, we hope someone pulls out the piece we released this morning and asks:
“were all the population assumptions wrong from the beginning”?
“if so, what does that look like today?”
See comparison of RCP8.5 (based on ~11 billion by 2100, circa 1990) vs. updated SSPs.
Uh, yeah.
https://envmental.substack.com/p/the-population-bombing
Have any of these temperature observations already been adjusted, particularly for UHI?
(I am an Aussie, more familiar with our own numbers than US numbers).
Roy, I shall be able to send you shortly a file of about 50 Australian stations that are plausibly the least affected by UHI, likely more pristine than some of your rural stations.
Geoff S
Randon useless thoughts . . .Environmental movement kicked in in the 60
s with time and money to be able to create inner city parks and green spaces and and plant trees and create fountains, and put sprinklers onto lawns etc ?always on / keeping the fire burningAnd enegy in the home has moved from
to light the gas / flick the switch only when required. Lighting moved from thermally inefficient kerosene and incandescent electric to fluorescent and LED ?Industrial parks / zonesare a relatively recent invention have very low population density but do some of the energy intensive stuff now in their own area rather than older smaller units within city/town boundary. ?There’s enough variability in the “standardized” Stevenson Instrument Shelter design, siting and maintenance procedure evolution to make the popular 0.0 plus pick a number per decade “questionable”. Example from below :”WMO 2010 recommendations, if incomplete, are a sound basis…”. Hello? 2010? 2010!,,,if incomplete?
Stevenson Instrument shelter (Wikipedia, abbreviated).:
… in 1884 included a double roof, a floor with slanted boards, and a modification of the double louvers.[4] This design was adopted by the British Meteorological Office and eventually other national services, such as Canada. The national services developed their own variations, such as the single-louvered Cotton Region design in the United States.[5]
The World Meteorological Organization (WMO) agreed standard for the height of the thermometers is between 1.25 and 2 m (4 ft 1 in and 6 ft 7 in) above the ground.
The interior size of the screen will depend on the number of instruments that are to be used.
The top of the screen was originally composed of two asbestos boards with an air space between them. These asbestos boards have generally been replaced by a laminate for health and safety reasons. The whole screen is painted with several coats of white to reflect sunlight radiation, and usually requires repainting every two years.
Siting[edit]The siting of the screen is very important to avoid data degradation by the effects of ground cover, buildings and trees: WMO 2010 recommendations, if incomplete, are a sound basis.[7] In addition, Environment Canada, for example, recommends that the screen be placed at least twice the distance of the height of the object, e.g., 20 m (66 ft) from any tree that is 10 m (33 ft) high.
In the northern hemisphere, the door of the screen should always face north so as to prevent direct sunlight on the thermometers.
A special type of Stevenson screen with an eye bolt on the roof is used on a ship. The unit is hung from above and remains vertical despite the movement of the vessel.
Future[edit]
In some areas the use of single-unit automatic weather stations is supplanting the Stevenson screen and other standalone meteorological equipment.[citation needed]
What about the changes in construction materials since 1800s, asphalting of dirt roads, brick & concrete replacing timber, etc ?
A thought inspired by the latest video from Jim Steele, right after he quotes you, at about 12:05, Steele brings up an infographic, intended for dog owners, illustrating the temperature differences between asphalt, concrete, gravel, and grass, with a ~50° F difference shown in his example.
Steele video link: https://www.youtube.com/watch?v=yknjuizJP6Y “The Science of Dryness and California Droughts and Fire”
Applied to your urban-rural temperature difference, I think there could easily be a correlation between the use of brick, tarmac (tarmacadam), concrete, asphalt for road and pad surfaces over the traditional use of dirt and clay, wood and corduroy road, grass, and cobblestone in suburban and rural areas (other than on highways and other finished trade roads). Asphalt and tarmac also saw beginnings of use on flat or low-pitched roofs, where tar was used as abinder, but also as a sealant and adhesive.
In the time frames you name, Coal tar was cheap and available. Coal and gas fired brick kilns made brick building and paving material more common. Farm equipment became too heavy for all-season, all-weather transport on unimproved roads. Mills, silo/storage, and train loading platforms, in or near areas convenient to telegraph and weather stations might be sited, were paved to increase efficiency and availability.
Even if only 10° F change of surface temperature was experienced, if close enough to a weather station, it should make a difference to air temperature readings.
Story tip:
A few things which might work together to explain this is the general increase in energy efficiency as the cost of energy increases, also if you look at population density from the point of view of persons per household, there is a downward trend. This has more of an effect in rural area where you can see the reults in urban sprawl. Although the topography and modern zoning laws have a lot to due with the extent of sprawl.
Late 1800’s urban air pollution levels!