April 28th, 2023 by Roy W. Spencer, Ph. D.
This is part 4 of my series on quantifying Urban Heat Island (UHI) effects on surface air temperatures as reported in the monthly GHCN datasets produced by NOAA.
In previous posts I showed results based upon monthly-average Tavg, which is the average of of daily maximum (Tmax) and minimum (Tmin) temperatures. Since late 2019, NOAA produces monthly average datasets for only Tavg, but since there are large differences in the UHI effects between Tmin and Tmax (urban warming is much larger at night than during the day, thus affecting Tmin more), John Christy wanted me to compute results for the older Tmax and Tmin datasets archived by NOAA.
As I have discussed previously, our computations of UHI are, I believe, rather novel since we do not classify stations as urban or rural. That is how most researchers have approached the problem. But as I have mentioned before, UHI warming occurs much more rapidly at very low population densities (PD) than it does at high population densities for the same population increase. As a result, a small population increase at a rural station can produce the same spurious warming as a large population increase at an urban station. This means that previous published results showing little difference between rural and urban trends did not actually demonstrate that homogenization methods actually remove UHI effects from temperature trends.
Instead of classifying stations as either rural or urban, we use regression to compute the slope of temperature-vs-population density in many sub-intervals of 2-station pair average population density, from near-zero PD to very high PD values. Then we integrate these regression slopes through the full range of population densities.
Since NOAA’s GHCN Tmax and Tmin dataset (v3) does not have nearly as many stations as their newer (v4) Tavg dataset, I have combined the 2-station matchups for May, June, and July rather than showing results for an individual month. I have used all matchups every ten years from 1880, 1890, 1900,… 2010 that are within 150 km and 300 m elevation of each other. All land stations from 20N to 80N latitude are included. I have computed results for both the unadjusted data as well as the adjusted (homogenized) data.
The results (below) show that the total UHI effect in summer for highly-populated stations averages 3.5 times as large in Tmin as it does in Tmax. Each curve is based upon approximately 300,000 monthly 2-station matchups.
The nonlinearity of the relationship is, as other investigators have found, very strong.
Note that the UHI effect shows up more strongly in the adjusted GHCN data than in the unadjusted data. I cannot explain this. It is not because of the weeding out of bad temperature data, because that only affects regression coefficients if noise is reduced in the independent variable (2-station population density differences), and not in the dependent variable (2-station temperature differences). The 2-station PD differences do not change between the raw and adjusted GHCN data.
As I have mentioned before, the above results do not tell us the extent to which GHCN temperature trends have been affected by urbanization effects. SPOILER ALERT: My preliminary work on this suggests UHI effects are rather large between 1880 and 1980 or so, then become quite small compared to observed temperature trends. But it must be remembered that here we are using population density as a proxy for UHI, which is not necessarily optimum. It is possible for UHI effects to increase as prosperity increases for a population density that remains the same.
Roy,
Australia provides a good test bed because of long duration data (some back to 1860s) and low population density. For many months now I have been trying to prepare a set of 50 Aussie “pristine” stations with Tmax and Tmin daily BOM temperatures and other observations like rainfall. The purpose is to create a pattern of baseline, weather/climate alone variations over 120 years in this part of the world.
I have selected the 50 stations and assembled adequate demonstration of lack of population effects. But there, it gets hard. This is mainly because of signal:noise limits in the data. Strange correlations appear, some being: long term trends by customary linear least squares fit are all over the place, from about -1 to +4 degC per Century for Tmax, (Tmin not all done yet); longer data sets (more years) show lower trends;
Trends correlate with distance from the nearest ocean; trends show some relation to station altitude; trends correlate with World Meteorological Organisation number.
There are known influences on trends that I am trying to remove to see residuals. These include metrication from F to C in 1975; change from liquid in glass to Platinum resistance in mostly 1996: change in screen size and shape now and then; station relocations; periods of recording to the nearest whole number C; data errors, such as cut and paste of blocks of data up to a month at a time; unusual data with strings of same consecutive numbers up to 7 days in a row; and more effects.
These raw pristine temperatures need correction for these effects. Some corrections are shown by colleague Dr Bill Johnston on Bomwatch blog, using corrections for rainfall that typically amounts to tens of percent of the total variation. If these effects were small enough to ignore, the 50 trends would match each other in magnitude and pattern. They do not.
In short, I am on the verge of dumping all this exploratory work on the unsuspecting scientific public, with a conclusion that the numbers, originally observed for purposes of early 1900s curiousity are not fit for the purpose of quantification of zero UHI.
Sadly, I suspect that your work will also find that, but at least we can say that we know why it failed, so that wheel needs no more reinvention.
Interestingly, the study I have done so far is entirely consistent with zero global warming at these 50 stations over the last Century.
Geoff S
Thanks for your work. It might be frustrating for you but these words are instructive for me. Another reason not to have much confidence in the grand narrative, whatever that might be. Future generations might get it all settled out but for now the evidence is mounting that being cautious about developing public policy predicated on suspect temperature records might not be very wise.
Dr. Spenser, thank you for all the fine work you and Dr. Christy have done over the years in advancing the science of weather and climate, especially atmospheric monitoring advances.
To protect my personal emotional wellbeing, however, I’ve ceased trying to make sense of all of the analyses and issues surrounding surface temperature measurements. It is my belief that poking at the enterals of official governmental land surface datasets and pre-ARGO SST speculations has no value in understanding the science relating to slowly evolving climate regimes all over the planet, except in approximating the grossest direction of global temperature trends.
Radiosonde records, satellite-based datasets (UAH6 and the new NOAA STAR) and ARGO seem to be the closest we can get to scientifically valid data on the evolving temperatures of the various parts of the Earth that matter; the bulk atmosphere and ocean strata. But my opinions carry no weight since I retired. In fact, I’m down to sneaking my testicals from my wife’s purse when action is needed.
Good report by Dr. Roy, as far as it goes. His comment suggests the need for a “prosperity” index. When a culture has more wealth they have central heat in the winter and central air conditioning in the summer, both of which add to the ambient temperature. How about bigger cars? Faster horses and more whiskey?
Also, Lawyers, Guns, and Money
vuk:UHI effect on the average temperature
ChatGPT:
One way that UHI can affect average temperature data is by creating a bias towards higher temperatures in urban areas. This means that temperature data collected from urban areas may be higher than the true average temperature for the region. This can be problematic when trying to compare temperature trends over time or across different locations.
To address this issue, some researchers have attempted to adjust temperature data to account for the UHI effect. This can involve using statistical methods to estimate the UHI effect and then subtracting it from the temperature data. However, this approach can be challenging as the UHI effect can vary depending on factors such as the time of day, season, and weather conditions.
Another approach to mitigating the impact of UHI on temperature data is to collect data from a mix of urban and rural locations. This can help to provide a more representative sample of the true average temperature for the region. In addition, some cities have implemented strategies to reduce the UHI effect, such as increasing green spaces and using reflective or lighter-colored surfaces for buildings and roads.
Make of it whatever you wish.
What I make of ChatGPT is you can’t trust it to tell you the truth because it makes things up out of thin air that have been proven to be ChatGPT falsehoods.
I wonder what would happen with ChatGPT if it were asked to write a story with an impossible paradox at its center? I’m thinking of Robert Heinlein’s story “By His Bootstraps”, where the protagonist (and practically the only character in the story) is himself at various ages in his time loop.
The paradox is that the protagonist (Bob Miller) is writing at his desk when a time portal appears in his room and a slightly older version of himself emerges and tells his younger self to enter the time portal and his fortune will be secured after he has a talk with an “old guy.” They have a few drinks while discussing it, and while they’re at it a third version of himself shows up. A fight starts, and Bob#1 gets knocked through the portal and Bob#1 meets the old guy.
This goes on through five Bobs, the “old guy” being the final version of Bob, and him realizing that all of the Bobs had been him, and that there was no place the time loop actually began.
Could an AI handle that kind of non-logical progression? Captain Kirk always managed to beat computers that way — how would the real deal handle it?
ChatGPT is supposed to be correlating existing data, so how does it go from doing that to completely making up false stories? The false story didn’t exist before ChatGPT got involved and created it. Why did ChatGPT create a false narrative? What would motivate it to do so?
I think the programmers have a lot of work to do. ChatGPT needs some tweaking.
“It is possible for UHI effects to increase as prosperity increases for a population density that remains the same.” Why not? Energy use is a commodity that is directly proportional to standard of living.
Where is one going to find an example of a place that has increased its standard of living while maintaining a constant population?
Population decline is one of Japan’s most significant issues – it has been falling since 2008 and is estimated to dip below 50 million by 2100
What strikes me as most interesting is how quickly Tmax peters out with poplation density. Tmin also peters out eventually. Since population density has been rising, but the effects peter out, one would expect that UHI contributes less and less to global average going forward.
The models have most likely been awarding to CO2 that which belongs to UHI. As time goes forward, they will diverge even more from actual temps as their over estimation of CO2 continues, while the factor that in part drove it (UHI) declines in effect.
It would be reasonable to suppose that the UHI effect in Sydney will indeed “peter out” (ie, go nearer to zero increase, not go nearer to zero), because the Sydney basin is full. But between Sydney and Canberra the amount of greenfield development is staggering. The new Western Sydney Airport and Wilton are only a part of it. It would be very reasonable to suppose that this part of NSW will step in and supply all the UHI that will not be supplied by Sydney.
As with all things climate, it is mostly the non radiative processes which we impact down here on Earth.
The urban environments increase boundary layer depth by several processes. Increased proportion of sensible heat and increase surface heat storage, namely.
A newly emerging urban development might start with lots of concrete, but low population density.
Increasing population density will correlate with the actual physical structure, by installing higher buildings. This physically impacts upon turbulent mixing. This will not increase the local temperature, but increases the plume of urban influence downwind.
The turbulent urban plume creates low capping inversions downstream, reducing latent flux and condensation even over rural environs by teleconnection. This depends on the actual physical structure.
“I have used all matchups every ten years from 1880, 1890, 1900,… 2010 that are within 150 km and 300 m elevation of each other. All land stations from 20N to 80N latitude are included. I have computed results for both the unadjusted data as well as the adjusted (homogenized) data.”
Roy, a scientific tour de force! That you obtained a family of such smooth related functions is self corroborative. That the adjusted data gives an exaggerated UHI result is forensic, suggesting that of a range of values in the homoge process, the largest values were deliberately selected.
Things that make you go hmmm…. DMI hasn’t updated their site on Arctic ice and weather since 22nd April.
Perhaps they don’t like the trend?
There are regular fails to update. Usually nobody at DMI has noticed. There is a contact email address on the site.
In my experience they fix within a couple of days,
This is still updating
https://ocean.dmi.dk/arctic/icethickness/thk.uk.php
I don’t see the purpose of averaging Tmax and Tmin. It tells us very little. The important information is Tmax and Tmin, their average is meaningless. How on earth can the adjusted temperature be higher than the measured temperature?
The “average”, actually a mid-range value, tells you nothing. When different Tmax/Tmin combinations give the same mid-range value how can it tell you anything? When temps like 80F/20F and 60F/40F both give you a mid-range of 50F what can you make of the 50F? One is likely a high desert springtime temp and the other springtime in the central plains of the US. Different variances in temps also means anomalies will carry the same variance, meaning the anomalies of the mid-range values are not as informative as many would have you believe.
Thank you! Let’s keep saying it: “the average of measurements is not a measurement!”
I’ve been trying to get this across for 2 solid years – and have failed with those in climate science. Ask yourself why they *never* propagate variance forward with their “average”. It’s because the variance gets so large that they can no longer claim they know the “change in temperature” down to the hundredths digit.
And when they compare averages in Las Vegas, a dry climate, with averages in Miami, a wet climate, they are comparing apples and oranges. They should be looking at enthalpy, not temperature! The data to do enthalpy has been available since automated temperature measurement stations (40 years?) but climate science refuses to change. They should also begin to use degree-days, the integral of the entire temperature curve, but they also refuse to change to this newer method even though agricultural science and HVAC engineering made the change at least a decade ago.
Their excuse? Tradition. They *all* sound like Teyve in Fiddler on the Roof! They refuse to accept that they can continue the old way alongside the new way. My guess is that they are afraid the newer ways will show up the problems with the old ways!
Well said! The “variance” in Earth surface temperature at any instant is at least from 233K to 313K, making +0.5K “change in average” seem rather inconsequential.
Yes! Tmax and Tmin are highly correlated. The average is contaminated from this. It makes any conclusions drawn from the Law of Large Numbers and/or Central Limit Theory invalid throughout the entire series of averages.
Tmax and Tmin should be evaluated separately. There is a 97% consensus that increasing Tavg means daytime temps are what is increasing. Climate scientists SHOULD be worried about this and endeavor to create a better understanding.
Why couldn’t it be increasing TMins just as well?
That’s the point! Increasing Tmin would not support the “Earth is burning” CAGW/CACC panic!
I did find out why Florida is heating up.
Are you being sarcastic, or just silly? Because:
a) airport weather stations are not primarily for climate analyses, rather they are for critical aircraft safety issues. (altimeter setting/atmospheric pressure; wind direction and velocity; visibility; cloud ceiling; and temperature/dewpoint are the main ones, in order of importance; critical for general aviation aircraft operation)
This information is available online, or in repeating automated radio broadcasts known as ATIS or AWOS services. The AWOS for K2IS is 124.175 MHz, and the METAR is:
https://metar-taf.com/K2IS
As a pilot, you check the METAR before you take off, say leaving Freeport in the Bahamas, and then again when you are say 20 minutes from your destination to see what the altimeter setting, wind, temp, visibility and cloud ceiling is. These critical aviation safety details are needed to be obtained on the airport grounds. Using this data for other functions like climate analyses is not the purpose that the aviation weather data is meant for.
And b) if you zoom out on your image, you find this teensy amount of paved runway and taxi ways, is in the middle of nowhere, surrounded by huge agricultural fields. Hence little to no UHI occurs.
And c) South Florida temps are not increasing! (I’ve lived here for 26 years)
According to the NWS … “The sensor should be at least 100 feet from any paved or concrete surface.”
To put the issue to rest, please find out if this weather station is used in climate modeling.
Regardless of opinions, this site does not meet NOAA standards. I asked the coordinator of the FAWN network, who said all FAWN observations are fed into the NWS data network … https://fawn.ifas.ufl.edu/
The amount of pavement is hardly the only impact. The amount of traffic is also a major impact. More traffic, more modulation of the temperatures read by the station.
That’s true for both airports and city streets.
“airport weather stations are not primarily for climate analyses, rather they are for critical aircraft safety issues.”
Absolutely true. So why are so many used by the climate science cadre?
Two points:
1) Yes, aviation weather stations are for aviation safety and must be located at airport grounds, and the actual temperature, etc. of air above the hot concrete or asphalt runways is a critical safety issue. So why are these data used in weather and climate analyses? My local TV weather persons always put a caveat on the temperature at the local airport KPBI (Palm Beach International), saying it is always several degrees warmer than actual air temp elsewhere…. (KPBI is 100x more area and concrete than K2IS)
While I wholeheartedly agree with Anthony’s crusade against improperly sited weather stations in general, aviation stations are correctly sited and should not be changed for aviation safety reasons. But they are not fit for general weather or climate purposes.
2) “Traffic is a major impact”….duh, do you know how much traffic is at this middle of nowhere airport? Well see the following:
http://www.airnav.com/airport/2IS
Scroll down to the section labeled: Airport Operation Statistics where it says “Aircraft operations: avg 32/day” for the year in 2018.
Now the “pavement” in question is the intersection of two taxiways and taxiing a plane is done roughly at engine idle speed… The likelihood of any of those 32 aircraft movements per day passing this intersection is perhaps at best 60% so 19 (small general aviation) aircraft pass this point per day, which we can assume is 80% daylight operation, so roughly 19 aircraft pass this point doing 10-15 mph taxiing in a 12 hour period. So 1.5 planes pass this point for maybe a few seconds each pass, per hour. With engine at idle, and with a 20-30 MPH propeller wash shooting back past the hot engine and exhaust diluting and dispersing the heat substantially.
Do you really think this amount of traffic affects the weather station readings? As Joe the big guy would say “C’mon man!”
“Traffic is a major impact”….duh, do you know how much traffic is at this middle of nowhere airport? Well see the following:”
“Do you really think this amount of traffic affects the weather station readings? As Joe the big guy would say “C’mon man!””
Yep, it just depends on when that traffic happens. If it happens at a time that would affect Tmax (e.g. increased air flow by the measurement station even at an engine idle) or Tmin then it definitely has an impact. Since Tmax and Tmin are the two temps used by climate science those are the only ones that need to be considered.
This wouldn’t matter nearly as much if climate science joined the 21st century and recorded the entire daily temperature profile at 1min or even 5min intervals. The integration of the temperature profile would minimize the impact of just those two temps on the daily mid-range value.
You didn’t answer my question: “So why are so many used by the climate science cadre?”
It’s my opinion that climate science would be far better off picking ONE station in the NH and ONE station in the SH and watching what happens at just those two measuring points. If what is happening is truly *global* then you don’t need a multiplicity of measurement points. Two would do just fine!
If what is happening is *not* truly global then why is it called a “global average”?
The Met office has been allowing 0.2 C for Urban heat effect since 1974 I think should be updated
0,1C per decade, which would now be 0.6 C but for very large urban areas, such as London even more.
reminds me of all the times, as a kid, i’d walk barefoot across our blacktop driveway on a sunny day in the summer. 50 years ago.
OT, but has anyone notices how cold ClimDiv says the US has been since 2005 🙂
National Temperature Index | National Centers for Environmental Information (NCEI) (noaa.gov)
And of course, you can see what is happening to USCRN… basically dead flat.
GHCN uses 10,000 pop as an upper limit for their Rural group.
The BoM in 2001 published the following paper setting out evidence of obvious UHI response in villages way under 10,000 population.
Torok, S.J., Morris, C.J.G., Skinner, C. and Plummer, N. 2001. Urban heat island features of southeast Australian towns. Australian Meteorological Magazine 50: 1-13.
http://www.warwickhughes.com/climate/seozuhi.htm
Page 7 at
http://www.warwickhughes.com/climate/seozuhi.htm
shows results of BoM temperature transects at Hamilton, Camperdown and Cobden in Western Vic that demonstrate higher temperatures in the very small villages Camperdown and Cobden.
Here is an experiment I carried out in 2010.
Two degrees C Urban Heat Island in small village of Barmedman, NSW, Australia 25May2020
http://www.warwickhughes.com/blog/?p=575
NOAA, CRU, Uni of Norwich, GHCN, GISS and others great and good have all lied to us for decades about UHI.
The big global temperature databases of the “great and good” are riddled with UHI.
“It is possible for UHI effects to increase as prosperity increases for a population density that remains the same.”
In that scenario, isn’t more likely that UHI goes down? As a given area, with constant population density, ages it would be likely that more vegetation and other cooling items develop (trees get larger, public parks developed, houses covered in ivy and roses, asphalt greyer/not as black as when new), so the UHI profile would level off and even drop slightly over time.
However you did say “prosperity increases” and not just over time in general – more and larger vehicles in the driveway (covering the asphalt), more swimming pools (I haven’t a clue or inkling if that would be more reflective or absorptive compared to a backyard full of grass and kids toys), even more cafes and patio restaurants with umbrellas and tents but also more decks that probably absorb heat better than the default grass or vegetation – makes for an interesting uhi lifetime profile.
It’s a shame that recent Tmin/max data has been imprisoned and not allowed to see the light of day or independent research.