I believe this is a truly important piece of work. I hope Dr. Spencer will submit it to a journal. I’m grateful to Dr. Spencer for his email suggesting I post it here. Consider this early peer review. Beat it up, find any errors, and point out flaws, so that he can make it better. – Anthony
by Roy W. Spencer, Ph. D.
UPDATED (12:30 p.m. CST, March 3): Appended new discussion & plots showing importance of how low-population density stations are handled.
ABSTRACT
Global hourly surface temperature observations and 1 km resolution population density data for the year 2000 are used together to quantify the average urban heat island (UHI) effect. While the rate of warming with population increase is the greatest at the lowest population densities, some warming continues with population increases even for densely populated cities. Statistics like those presented here could be used to correct the surface temperature record for spurious warming caused by the UHI effect, providing better estimates of temperature trends.
METHOD
Using NOAA’s International Surface Hourly (ISH) weather data from around the world during 2000, I computed daily, monthly, and then 1-year average temperatures for each weather station. For a station to be used, a daily average temperature computation required the 4 synoptic temperature observations at 00, 06, 12, and 18 UTC; a monthly average required at least 20 good days per month; and a yearly average required all 12 months.
For each of those weather station locations I also stored the average population density from the 1 km gridded global population density data archived at the Socioeconomic Data and Applications Center (SEDAC).
All station pairs within 150 km of each other had their 1-year average difference in temperature related to their difference in population. Averaging of these station pairs’ results was done in 10 population bins each for Station1 and Station2, with bin boundaries at 0, 20, 50, 100, 200, 400, 800, 1600, 3200, 6400, and 50000 persons per sq. km.
Because some stations are located next to large water bodies, I used an old USAF 1/6 deg lat/lon percent water coverage dataset to ensure that there was no more than a 20% difference in the percent water coverage between the two stations in each match-up. (I believe this water coverage dataset is no longer publicly available).
Elevation effects were estimated by regressing station pair temperature differences against station elevation differences, which yielded a cooling rate of 5.4 deg. C per km increase in station elevation. Then, all station temperatures were adjusted to sea level (0 km elevation) with this relationship.
After all screening, a total of 10,307 unique station pairs were accepted for analysis from 2000.
RESULTS & DISCUSSION
The following graph shows the average rate of warming with population density increase (vertical axis), as a function of the average populations of the station pairs. Each data point represents a population bin average for the intersection of a higher population station with its lower-population station mate.
Using the data in the above graph, we can now compute average cumulative warming from a population density of zero, the results of which are shown in the next graph. [Note that this step would be unnecessary if every populated station location had a zero-population station nearby. In that case, it would be much easier to compute the average warming associated with a population density increase.]
This graph shows that the most rapid rate of warming with population increase is at the lowest population densities. The non-linear relationship is not a new discovery, as it has been noted by previous researchers who found an approximate logarithmic dependence of warming on population.
Significantly, this means that monitoring long-term warming at more rural stations could have greater spurious warming than monitoring in the cities. For instance, a population increase from 0 to 20 people per sq. km gives a warming of +0.22 deg C, but for a densely populated location having 1,000 people per sq. km, it takes an additional 1,500 people (to 2,500 people per sq. km) to get the same 0.22 deg. C warming. (Of course, if one can find stations whose environment has not changed at all, that would be the preferred situation.)
Since this analysis used only 1 year of data, other years could be examined to see how robust the above relationship is. Also, since there are gridded population data for 1990, 2000, and 2010 (estimated), one could examine whether there is any indication of the temperature-population relationship changing over time.
This is the type of information which I can envision being used to adjust station temperatures throughout the historical record, even as stations come, go, and move. As mentioned above, the elevation adjustment for individual stations can be done fairly easily, and the population adjustments could then be done without having to inter-calibrate stations.
Such adjustments help to maximize the number of stations used in temperature trend analysis, rather than simply throwing the data out. Note that the philosophy here is not to provide the best adjustments for each station individually, but to do adjustments for spurious effects which, when averaged over all stations, will remove the effect when averaged over all stations. This ensures simplicity and reproducibility of the analysis.
UPDATE:
The above results are quite sensitive to how the stations with very low population densities are handled. I’ve recomputed the above results by adding a single data point representing 724 more station pairs where BOTH stations are within the lowest population density category: 0 to 20 people per sq. km. This increases the signal of warming at low population densities, from the previously mentioned +0.22 deg C warming from zero to 20 people per sq. km, to +0.77 deg. C of warming.
This is over a factor of 3 more warming from 0 to 20 persons per sq. km with the additional data. This is important because most weather observation sites have relatively low population densities: in my dataset, I find that one-half of all stations have population densities below 100 persons per sq. km. The following plot zooms in on the lower left corner of the previous plot so you can better see the warming at the lowest population densities.
Clearly, any UHI adjustments to past thermometer data will depend upon how the UHI effect is quantified at these very low population densities.
Also, since I didn’t mention it earlier, I should clarify that population density is just an accessible index that is presumed to be related to how much the environment around the thermometer site has been modified over time, by replacing vegetation with manmade structures. Population density is not expected to always be a good index of this modification — for instance, population densities at large airports can be expected to be low, but the surrounding runway surfaces and airplane traffic can be expected to cause considerable spurious warming, much more than would be expected for their population density.
If UHI is largely caused by man-made surface materials (concrete, black top, shingles, etc.), don’t we need to get an estimate of the surface area that is covered by those materials for the area surrounding the station? Also, what about average humidity levels?
Software plus decent satellite imagery would be a good start for generating this kind of metadata. Has anyone attempted to do this yet?
Dr. Spenser said: (13:20:43)
“Keep in mind that these are the average UHI effects across ~10,000 stations. Obviously, the UHI effects at specific stations could be much larger.”
This hits on what I’m wondering – what you’ve described is a way to adjust for UHI, but not problems with proper station siting, if I’m understanding correctly.
Further corrections are needed for the siting problems.
Personally, I do not like estimating temperature by guessing about it.
Trying to guess the effect from a certain amount of people, etc.
Some of the national parks have long histories of temp, why not start there?
Nice work. The study seems to suggest that land-use change (at least initially) has a higher impact on temperature than actual energy use by cars, industry, heating, etc.
Perhaps there is also some sort of tipping point (ugh!) in the UHI-effect, where further urbanization actually has a cooling effect, f.ex through shading effects from tall buildings, funneling effects on wind, etc?
Just speculating..
It should be clarified what you mean by the pairs in which BOTH locations are at the lowest density category. It sounds as if there is no spread in population in these pairs, which would of course mean that you can infer no effect in terms of UHI. But I presume you mean that the bottom bin, 0-20, has been further divided (as is possible from the population data you’re using) and temperature differentials between pairs from different levels in this bin are extracted.
I believe your tentative language is wise; this is a nice looking piece of parlor science, but as you say before trying to make any bold assertions about it it would make sense to repeat the procedure with data from different years to demonstrate robustness. Still, with the size of the data set you can infer a pretty strong statistical correlation even as it stands. I agree with Anthony that this should be published with high priority, but I will agree with your implied intentions, that one should first establish robustness of the effect by varying your sample.
Also, I suspect that the population density statistics are as flawed, or nearly as flawed, as the temperature data sets, the only difference being that they are spoiled by fewer and less-well-paid cooks, and so of course are somewhat less spoiled.
How sensitive is the result to the maximum spacing, 150 km? Is it much different is 100 km is used? or 250 km?
As long as you keep yourself to way back when like the year 2000 you gonna get good enough population density statistics, but the more resent they are the more sketchy they are. Even just going through the EU data you have statistics from different time periods mixed in and presented as updated until you read the fine print. For cities like London, New York, and Tokyo, the year to year difference in pretty minute, but for cities in developing countries especially China you can have up towards 1 million people added per year for ten years straight and if the data is from 2003, or 2005, or what ever but presented as the most updated data 2010 you’ll have a difference in density in some cities that number upwards eight million people in the same municipality. The statistics from Africa is even worse where the most, fairly old, but accurate statistics comes from African Union and UN and the refugee camps, and the cities tend to grow a lot when most of the warring is over, but theirs not exactly a priority of keeping track by how much. In South America, or Mexico and below, poverty pretty much screws things up in the statistical department, just look at Mexico City or Rio de Janeiro, and they are supposed to be quite stable in their population growth.
So take care in choosing your years, and make sure to only choose years where there’s an actual official number for, lest you gonna fuel the fanatics, considering how they’re lashing out at people for spelling mistakes and all.
Dr. Spencer,
Thank you for a fine effort. Several people have asked for more information about station pairing and some have asked about differences in socio-economic status of the regions of paired stations. If all stations pairs are within 150 km of each other, most are probably somewhat similar socio-economic regions. For example, the difference between US and Canadian stations would not be significant. However, the difference between US and Mexican stations could be. Might I suggest a change or clarification in the pairing so the stations are paired within the same country or at least within the same socio-economic range?
I expect that there will also be UHI variances as the temperature changes through the range where heat pumps and swamp coolers are effective. Heat pumps lose effectiveness at points that vary with the efficiency of the unit as the temperature goes down, and I don’t know the current range, but believe they will start being switched off between 20F and 10F as the temperature goes down. Thus the UHI from nearby heat pumps would accelerate as the temperature increases over about 75F as more A/C units switch on, and have an odd shaped curve as the temperature decreases through the heating zone until they turn off at around 10F – 20F on the way down. There would still be building heat loses below the cutoff point, but these would be much less than the compressor output.
Heat pump effect would be limited to areas that can afford them.
This looks really cool, but I do have a couple concerns about using this to extrapolate distant historical data (past 100 years or so).
1) Buildings may have changed both in materials used and in their shape (brick, glass, wood, concrete, etc).
2) Energy use has almost certainly changed over the past 100 years with different technologies and how widespread frequently they are used (computers, air conditioning, heating, lighting, etc).
I could go into specifics but I’m sure you get the point. That being said, it seems like UHI may be a bigger factor in the global record than many scientists admit – something I’ve suspected for quite some time.
Sorry Roy, agree with JDN. The description of what you have done is largely incomprehensible.
To JDN. From what I can gather, the ‘increase in population’ is between station pairs, not a period of time. (other than that clarification I’m lost)
A couple of pages, Roy, on what the following actually means would help a lot:
“The following graph shows the average rate of warming with population density increase (vertical axis), as a function of the average populations of the station pairs. Each data point represents a population bin average for the intersection of a higher population station with its lower-population station mate.”
Dr. Spencer,
Thank you for the comment just above. I understand the main concern is not the sensible heat from people’s bodies, but how do we know this is not significant? When you have a bunch of people in a room, you have to turn the a/c on. Has some study looked into the question and determined the sensible heat from bodies is not significant? If so, I would love to read the study.
In your last post Dr Spencer I suggested something of the same thing. Stop trying to avoid UHI and instead quantify it. I suggested concentric rings of weather stations from city centre to outskirts that would allow the UHI signal to be quanitifed and separated from the “natural signal”.
I think your station paring is a similar approach but I think it is too thin because it doesn’t give you enough data on the UHI profile. I would think that UHI profile would vary with prevailing wind, direction (north vs south of the major centre) latitude (larger amplitude of absorption of insolation in high versus low latitudes) the major economic driver of the city ( a financial centre, a manufacturing centre and a transportation centre all grow up very differently). So I think the technique is valid, I just think that two station pairings are insufficient.
(larger amplitude of absorption of insolation in high versus low latitudes)
should have read
(larger amplitude of absorption of insolation by vertical surfaces such as building in high versus low latitudes)
Dr. Spencer :
wayne (13:53:20) :
Seems I misread the exact meaning and implication of your Station Warm Bias vs. Population Density graph. The comment I made above addresses a very similar and related graph, Station Temperature Anomaly vs. Population Density which was not shown. Could you graph that also?
Great work. Following is some brainstorming:
The Urban Heat Island effect (UHI) includes:
1) Change in albedo affecting the radiative balance (fields vs asphalt).
2) Spatial density of energy use (heating from nearby air conditioners, heaters or jets)
3) Cumulative heat by convection from upstream areas.
2) Since you are using international data, the spatial density of energy use may show a correlation with per capita GDP for the local population.
3) The cumulative heat from upstream convection may correlate with the spatial energy density 2), the radius of the population density, and the wind speed. The upstream population is likely to scale with the population density.
Note the fractal correlations of population size. See: Fractal dimension and fractal growth of urbanized areasShen G.1, International Journal of Geographical Information Science, Volume 16, Number 5, 1 July 2002 , pp. 419-437(19)
Ron Cram,
I don’t think body heat is much of an issue, since people are dispacing deer. 😉
It might create a noticable effect when you squeeze a couple thousand penguins together, though.
Anyway, for a good science-fair experiment, spreading a hundred or so remote sensing/WiFi thermometers across an area that included a mix of meadow, forest, pavement, and housing woud produce some interesting numbers.
Correlation is not causation.
Can anyone name a mechanism that would cause less densely populated areas to warm more? Or is it simply that the less densely populated stations are located in the areas of the globe that are for other reasons undergoing more warming (e.g. the poles) and at the same time are less hospitable to human occupation (hence lower pop. densities).
I’d like to see these numbers broken out, for instance, by latitude bands, or by urban proximity (e.g. low-pop-stations within X km of any dense-population-station vs. low-pop-stations beyond X km of any dense-population-station). What if the stations showing the most warming are in the deserts, tundra and ice caps, and that the larger warming is thereby a reflection of their natural climate rather than anything having to do directly with population? That is, the people aren’t the cause of the warming, but rather the cause of the warming is also the cause for the lack of people.
I’d also like to see these numbers according to population growth rather than static density. The UHI argument as I understand it is that 30 years ago there was nothing there, and today their a parking lot and an air conditioner vent, and that makes readings now seem warmer than they really are. So what does current population density have to do with this? If population density has been >2000 ppl./sq. km. for fifty years, how does that imply spurious warming in that 50 year span?
I’m confused… right now this looks like playing with numbers, without a clear mechanism being defined or studied behind the statistics. It needs a lot more work.
I’ll bet GDP/area has tighter correlation with UHI than population density. Just a thought.
Jeremy (13:33:33) :
@Jan Frykestig (11:39:22) :
—> “” Spencer: “this means that monitoring long-term warming at more rural stations could have greater spurious warming than monitoring in the cities.”
One would then reasonably expect to see much higher trend at rural stations than at urban ones, ok? However, the USA data show exactly the opposite – just 0.1 deg C of warming in 20th century, but 0.6-0.7 degrees C in the cities. How that can be, if the rural stations should have a larger warming bias? “”
No. Spencer is talking 2nd derivative and you’re talking first derivative. Spencer is saying that you may see greater change in temperature anomaly at sites where the population is increasing from near zero to 1000 per square km. You are talking about greater temperature anomalies in the same areas. Do you see the difference? Spencer is saying depending on site, you might expect a rural station to show much greater change in anomaly per with respect to population changes, and you are asking why we don’t see greater temperature anomaly with respect to population. That is a mismatched question because Spencer is simply talking about a different slope than you are asking about.
======
But have we seen a larger rate of temperature increase in rural stations?
“Significantly, this means that monitoring long-term warming at more rural stations could have greater spurious warming than monitoring in the cities.”
What I would like to see is a simple average of the real raw data from all truely defined rural stations that have continious data from 1943 to the present.
This would not be gridded or manipulated in any way.
It would not be a global average, but a representative sample (if 100% is a sample) of how temperature has changed over the last 65 year zig-zag climate cycle.
If that was too expensive, then a properly chosen statistical randomised sample would be quite accurate, as long as done by someone with appropriate statistical qualifications and a reasonably open mind.
If too impractcal, then perhaps a proper random sample from Anthony’s rural USA data base would sufice.
If we can establish what has happened to temperature in areas with minimal UHI change, then we will know what is happening to global temperature and why.
With due respect to Professor Spencer, this seems a much quicker, simpler and direct way of going.
I suspect that the reported temperature increase since 1880 is due to the combined effects of UHI and data manipulation. I may be wrong.
Someone with access to the data and sufficient knowledge should be able to bring this whole circus to a head in a few weeks.
We will then know if we are facing a climate disaster or a politically driven, economic one.
David (12:57:12) : “Airport sations could be separated from other stations, and analysed against (for example) km of runways.”
Or maybe number of flights daily.
So let me see if I have this right, UHI produces a warming based on population density, check, makes sense — So then that ‘warming’ should be subtracted from the measured temperature to get the ‘real temperature’. T(real) = T – UHI(T) — UHI biases are always warmer than actual temperature, right?
So what happens when we get to the Chinese who continuously move their temperature stations away from airports and populations?
It’s beginning to seem like an impossible problem to solve with one set of relationships.
I’ve been lurking here for a long time, often impressed by the quality of work from Dr Spencer et al. But this time I hear alarm bells ringing: Can a population of 20 per sq. km really raise the temperature by almost 0.8 degrees C? A quick check would be to compare the latent heat of a sq km of atmosphere close to the earths surface with the heat output from 20 people.
With a cp = 1.0 kJ/kg K, an air volume of 1 cu. km and a density of rho = 1.2 kg/m3 and a temperature rise of 0.8 degrees we need almost 1 Terajoule of energy! How fast does the atmosphere loose heat to space?
” Ivan (12:51:10) :
I really, really cannot understand what is reason for all of these mathematical speculations. Is not far easier to select ALL RURAL stations in the USA and compare thus obtained trend with the urban trend? Is that idea really so unimaginable and stupid?”
Keep asking the question Ivan. Maybe someone who has the raw data will do this. What is there to lose? Let’s us just be told the rural trend.
I’m currently trusting no one. But the truth is still out there! Give us the data. Hypotheses can follow.