Over on Warwick Hughes’ blog, a new article by Lars Kamél on the GISS adjusting temperatures for an apparent “urban cooling effect” that mysteriously happens to some small Swedish towns but not others nearby.
I wonder how deep this “adjusting key data for unlikely causes” rabbit hole actually goes…
Climate Audit 
49 Comments
The village in question is in Gotland – see some tree ring information from Gotland here http://www.climateaudit.org/?p=360 . I’ve noticed some articles by Kullman on Swedish trees in which he questions the instrumental data.
I made a graph comparing mean annual temperatures for Sodankylàƒ⢠in northern Finland with data from GISS and NordKlim sets (see comments to Kamel’s post). Please check that I did not do anything stupid, since I’m rather amateur with climate data. Thanks!
It is just so unbelievable that after all this time since 1988 we do not have a surface temp dataset that is open for analysis and replication that can be used to give a measure of average global temp.
Am I wrong about this? Is anyone working on it?
#2: They’re just making it up at GISS. http://data.giss.nasa.gov/gistemp/2005/
I put a 1200 km circle around where I live, Boston MA. It includes the lower end of Hudson’s Bay and the Island of Bermuda. Polar bears and tropical fish living in happy harmony. It would be very interesting to know which weather stations are the basis of the estimated temperatures. Of course, there’s the very obvious point that for this to be science, you’re supposed to measure data not guess a number. My bozo meter pinned at 11 when I read the above quote.
If you look at the JunkScience.com compilations of surface temperatures, GISS is consistently higher than any of the other estimates of the surface temperature, which in turn are all higher than the satellelite measurements.
Hi folks,
I spent a few minutes perusing the GISS data for areas I am familiar with, and found a couple of odditities. Please note, this isn;t cherry picking, they were simply the first two sites I looked at, as I had growm up in one (Auckland NZ),and currently live at the other (San Francisco CA)
In the population statistics, which is presumably used to adjust for UHI, the Auckland figures were all 145,000. This is odd since the population of Auckland is closer to 1m people. Perhaps the population is divided out over the area?
Then checking San Francisco, all the sites around here were showing 6.25m people. This might be accurate for the Bay Area as a whole, but seems excessive for semi rural sites, like Hamilton/Afb. Other sites in the area seem better represented.
This is an area where tracking the population changes along with temperature readings would be fruitful.
Sadly there are thousands of sites, and collecting and collating that data is a monumental task.
cheers,
Robert.
I’ve always wondered how they correct for the urban heat island effect.
It seems like the most straightforward way to do this is to compare the trend at the urban site to the trend at a nearby rural site and correct the urban trend to be equal to the rural trend.
But, if you do this, the urban trend becomes worthless for assessing temperature trends, and so trends should simply be measured by throwing out the urban observations.
What am I missing? How do they actually correct for the UHIE?
Steve Milloy pointed me to the NCDC for global absolute mean temperature data which I plotted up here http://lhcrazyworld.blogspot.com/2006/02/absolute-global-mean-temperature.html.
Later posts looked to 1976 to 2005 and finally CO2 vs the 1976 -2005 data.
As far as I am concerned its global mean absolute temperatures that are the metric, not the temperature anomalies which, for representation of the data in a graph is nothing more than an enhancing technique.
The NCDC data shows from 1880 to 2005 a decadal increase of 0.8 deg Celsius.
Apologides a Correction – 0.08 degrees c per decade.
Oh dear, make that 0.051 deg C per decade, the 0.08 referred to 1976 to 2005.
Sorry Steve, doing two things at once. 🙂
Re: 6
Terry,
That is a very good question. The claim is made that the CRU data is corrected for UHI effects. The problem is that Phil Jones will not release his data so that we do not know how he did the adjustment.
There is another concern about the instrument data. Many stations have been closing over the past 30 years. The current number is roughly half of the number of stations which were operating in 1970 (GISS data home page). I recently looked through the GISS list of stations. It appeared to me that the majority of recently closed stations were rural. If indeed, the station data set is becoming less rural than it was prior to 1980, then the proper UHI correction is more important today now than it was 25 years ago.
If the character of the GISS and CRU data sets is becoming increasing urban, then even a slight undercorrection for the UHI would allow urban climate to be mistaken for global warming.
Re #10,
I fully agree with that remark. While looking for rural stations in North Russia, of the six rural stations only one remained after 1980. What is left are large towns like St. Petersburg and Moscou. Hardly correctable for UHI.
Even worse is the situation in the equatorial band (20N-20S) where there are practically no reliable rural stations at all.
Look e.g. to the data for Salvador, a town of 1.5 million inhabitants. That should be compared with rural stations to correct for urban heat island effect. But the nearest rural stations are 458-542 km away from Salvador (Caetite, Caravela, Remanso). And their data are so spurious, that it is impossible to deduct any trend from them. Quixeramobin is the nearest rural station with more or less reliable data over a longer time span, and shows very different trends than Salvador. Or look at Kinshasha (what a mess!), 1.3 million inhabitants, Brazzaville (opposite the Congo stream), and something rural in the neighborhood (Mouyondzi – 173 km, M’Pouya – 215 km, Djambala – 219 km,…). East Africa is not better: compare the “trends” of Nairobi with these of Narok, Makindu, Kisumu, Garissa,… Rural data trends with some reliability on a longer time span are very rare in the whole tropics. Only expanding towns have (sometimes) longer data sets which are hardly correctable. The unreliability of the data in the tropic range is thus obvious, that one can wonder how a global surface temperature trend can be calculated to any accuracy…
re 2:
Adding an explanation as first sheet in http://www.smhi.se/hfa_coord/nordklim/data/Nordklim_data_set_v1_0_2002.xls
would be handy, don put in in aseperate xls file, they tend to live separate lives.
re4:
Annual temperature anomalies are well correlated within a range of 1200 km. Polyakov, finds the same correlation distance in the arctic, and I find the same in central Europe.
http://home.casema.nl/errenwijlens/co2/europe.htm
Re 10:
There has to be some explanation of how it was done and some discussion of that method. I have heard that there are multiple interpetations of the data and so, presumably, multiple meyhods of correcting for the UHIE. Somewhere there must be a summary description of these methods, a comparison of the relative merits of each, and at least a small debate about which is best.
re 12: I’m not sure if I understood your comment correctly. I do NOT have anything to do with NordKlim project, I just pointed out that data is available from there. NordKlim was a joint project of Finnish, Swedish, Norwegian, Danish, and Icelandic meteorological institutes, the ones who have all data here. Since any data (including GISS) has to originate from these institutes, I think it would be interesting to compare data available from GISS to the “original” data. Furthermore, the data set contains some very interesting stations (since they are up in north, and time series should be rather long in rural areas), see the map in the report.
What comes to my graph, I just took the average of all months for each year for Sodankylàƒ⢠mean monthly temparetaure (sheet 101) divided by 10. Since the resulting graph agrees with the one in Figure 6 in the report, I’m pretty sure I plotted it right. Sodankylàƒ⢼/a> is a completely rural station (current polulation 9300 with 0.8 inhabitants/km^2).
Re #10 and #13
I googled a bit to try to find how these UHI adjustments are made but all I found was William Connolley on RealClimate telling Ferdinand Engelbeen (21 Dec 2004 @ 4:10 pm) “You really need to read the Peterson paper more carefully. His assertion (and indeed Parkers) is that the series *don’t need any correction*.”
I find this a bit odd as one of the leading alarmists, DEFRA Minister Micheal Meacher, told my MP in December 2002 “The data are very carefully quality controlled and urban heat island effects have been removed”. Why would they be removing things if they aren’t there?
I asked the new Minister Elliot Morley if he would make the raw and adjusted data available for public scrutiny in November last but have had no reply.
Phil Jones was the author of a UHI study arguing that the effect was negligible. I asked him for the data for his Nature paper. He said it was on a diskette somewhere and could no longer locate it, but they had “moved on” from using that data.
I once asked CRU to provide me with a set of two temperature series of their choosing that showed the differences between the an uncorrected and UHI corrected temperature series. They refused. I think they make no correction for UHIs at all and assume it is small. Warwick Hughes can provide more information. One justification for their assumption is the work of Parker.
In a key statement, Parker says: “Here we show that, globally, temperatures over land have risen as much on windy nights as on calm nights, indicating that the observed overall warming is not a consequence of urban development.”
A better interpretation of Parker’s statement is that the urban heat islands are so strong that even windy conditions are not sufficent to carry a significant portion of the heat away and reduce the urban heat island effect.
See http://climatesci.atmos.colostate.edu/2006/01/29/do-urban-areas-have-larger-long-term-temperature-trends-than-other-locations/ for more discussion.
It is hard to believe this is the basis for ignoring UHIE. I thought the “windy day” paper was very recent while the temperature record and UHIE research has been around a very long time.
The IPCC 2001 TAR says, (in section 2.2):
Section 2.8 says:
The problems arise when you use gridded data to calculate a “global average” temperature. Many grids have seen a reduction in rural stations and are thus influenced by urban stations far more than they were prior to 1970. Because of this, the effect of UHIs is increasing. If you look through the GISS data station by station, you will find that the graphs of urban stations resemble Mann et al’s Hockey Stick whereas rural stations generally do not.
I found Parker’s paper interesting, however I disagree with his conclusion. Cities cool at night by both radiation and convection, therefore, it would not be surprising to have nighttime temperature inversions over urban areas. On windy nights, I would expect that the air over a city would mix much more than it would on calm nights. Mixing would reduce the effect of a temperature inversion, leading to warmer temperatures on windy nights. In my opinion, Parker’s paper does not show that the UHI effect is neglible, it actually shows that it exists.
“In all cases these trends are positive. The increase in the UAH time series is 0.12°C/decade (0.22°F/decade), 0.14°C/decade (0.24°F/decade) for the RSS analysis and 0.10°C/decade (0.17°F/decade) for the University of Washington. Trends in UAH, RSS and UW data are less than the trend in global surface temperatures, which increased at a rate near 0.18°C/decade (0.32°F/decade) during the same 27 year period.” From http://www.ncdc.noaa.gov/oa/climate/research/2005/ann/global.html.
The satellite observations give a warming of 0.12 +/- 0.02 C/decade based on 3 analyses. Assuming for the moment that all of it is caused by greenhouse warming, the greenhouse warming theory predicts it will be about twice as strong in the mid-troposphere as what will occur at the surface implying 0.06 C/decade warming at the surface would arise from greenhouse gases. It is observed to be 0.18 C/decade, so it seems a warming of 0.12 C/decade must be arising from some other cause other than greenhouse gases. Potential causes could be UHIs, land use changes, or changes in cloud cover, for example. If it is UHIs, then they are contributing 0.12 C/decade to warming or two-thirds of the measured surface warming is spurious.
Another way to look at it would be consider formulas by Oke or Torok that show urban heat islands contribute to the observed warming in proportion to the log of the nearby population. The amount of land is finite and world population has increased by about a factor of four since 1900 so, on average, one would expect the population density at a typical station would increase by a factor of four as well. The expected typical increase from this factor then would be about 1.5*log(4) or about 0.9 C which is even larger than the observed warming. It amounts to 0.09 C/decade over a century, close to the figure above. It seems to me that UHIs may very well be playing a major role in contaminating surface temperature measurements.
Another concern (to me at least) is reporting of stations. For those that follow junkscience I’m sure you aware of the large loss of many reporting stations on December 5th
http://www.junkscience.com/GMT/Change/Change.htm
As they say there, not saying that there is a conspiracy at all. But is the fact that many of the northern stations will stop reporting in the winter due to seasonal ceasing of activities. A semi-rural airport in the far distant North West Territories isn’t going to stay operational through the winter months, so they shut down and their data is lost, till at least they resume operations. This isn’t something your going to see happen in Aruba. As a result is the loss of northerly reporting stations a regular occurrence? does that skew temperatures as a result of loosing one tail of the bell curve? Has this always been the case? Most importantly is it taken into consideration?
Just yet another reason that taking a global mean temperature is so difficult. Ideally you need an even distribution of similarly accurate stations, that’s perfection and we will never reach that with ground based monitoring. As a result satellite temperatures (that do not show the warming that ground based stations do) should be given higher confidence levels.
Re #21
Doug,
I used this data to calculate the absolute global mean temp for my post #7 above.
ftp://ftp.ncdc.noaa.gov/pub/data/anomalies/global_meanT_C.all
This data contradicts all the decadal rates for the “temperature anomaly” graphs referenced by you. I got + 0.051 deg Celsius per decade from extrapolating linearly. The higher rates must therefore derive from modelling? Or is it the rate of the increase of the temperature anomalies?
Louis,
You are referring to 1900-present surface data and I was referring to 1979-present satellite data and then to modeling of trends, so there is no contradiction.
Re #21 Douglas,
To follow through on the logic trail in your post I would then have to conclude that if all the surface warming is not induced by tropospheric warming then all the tropospheric warming could not be the result of anthropogenic greenhouse gas as any surface warming should result in atmospheric warming. This leads to a diminishing return on the AGW component of the overall warming.
Re: 22
Seasonal shutdown of rural stations, particularly those located away from the Equator, makes certain grids more urban in character. These grids would indicate a higher winter temperature than would be the case if these rural stations continued to report. Therefore, not only should temperature data be adjusted for the UHI effect, but it now seems that some data should be adjusted for the effect of seasonal shutdown. Since both of these effects will increase the average annual temperature calculated from unadjusted data, it makes access to the data used to calculate global average temperatures even more important.
Dr. Jones, please step forward and show us your data!
#12:
That may be, but the point of my comment, which I didn’t express well, is that the temperature measurements are the data that are supposed to explained by the climate models. Once you start doing things like this you are inserting a model into what is supposed to be the data. In this case he is assuming 100% correlation between locations. What other assumptions and extrapolations are added that we don’t know about? What is the validity of these assumptions? With all this going on, the data have become the “data modified by modeling assumptions.”
The rule of experimental science has always been “know your equipment.” Know what it can and cannot measure and the errors that it introduces. You are allowed to correct for these when you present the results of your measurements with a full explanation of what you did. What you are not allowed to do is invent data but that is exactly what is being done here by extrapolating weather over 1200 km.
re 27:
Paul, You are right wrt to the extreme GISS results in the arctic: That’s an extrapolation on the edge of a grid, which is a no-no in aeromagnetic processing!
# 24
Doug,
right, and I did get a slightly higher trend in my data for the same period two. Thanks.
# 28
Hans,
unusual that the GISS produces an anomaly in the Artic which Spencer and Christy also show from their satellite data.
This suggests that it might not be a computing artefact unless they use the same algorithm, and if so, an artefact it is.
I am away until 10th but here are some quick comments,
First, generally re Urban Cooling in Sweden” check out my 9 Jan post re How NASA GISS inserts warming into USA rural T data;
http://www.warwickhughes.com/blog/?p=27
Read carefully Jim Hansen’s own words.
re #5 Robert, have a look at my page;
http://www.warwickhughes.com/climate/easterling.htm
I used figures from the website;
http://www.gazetteer.de/home.htm which is now at
http://www.world-gazetteer.com/
to specify many examples where GHCN stated populations are wildly inaccurate.
Re #6, Terry.
Jones et al 1986 and Jones 1994 and any other iteration, NEVER “correct for the urban heat island effect”.
Read my 20th Anniversary Review of Jones et al 1986 at;
http://www.warwickhughes.com/cru86/
Look for the to and fro comments between Wood 1998 and Wigley & Jones in same journal for illuminating glimpses into what they did.
http://www.warwickhughes.com/cru86/wood.htm
Also note my 14 Dec Blog post re “How did Jones et al 1986 and Jones 1994 select Atlanta ?”,
http://www.warwickhughes.com/blog/?p=21#comments
see 21 Dec comment there re my Open Letter Number 1 to Jones et al 1986 on inclusion of Atlanta GA;
http://www.warwickhughes.com/opletts/
Note how Phil’s reply only says that “they used other stations too”.
He does NOT ever say, we adjusted Atlanta (or anywhere else) for UHI.
Re Terry’s #10, I know many of the great and the good (and the IPCC) make the claim “that the CRU data is corrected for UHI effects.”
I do not believe those claims would stand up to determined examination.
Best wishes, sorry for rush,WSH.
One should be very careful in trying to extract the UHI by comparing a rural station with a non-rural station. J. Christys work on the temperature stations in a valley in CA clearly shows that land use changes (e.g. farming) can give a temperature increase similar to that of cities (all stations in the valley showed dramatic worming the last 100 years – while the foothill stations showed no warming trend).
So if this effect is significant – bothe areas will show a similar trend. Then the UHI correction done by GISS or Hadley (Jones) is wrong since they both show an a similar increase (growing city and changes in land use).
Re # 32
Given enormous amount of adjustments to the raw data to ‘normalise’ them, according to preconceptions of what constitutes an anomalous warming due to anthropogenic effects, suggests that land based temperature estimates are not at all a useful metric for determining the earth’s GMT.
The ensuing statistical arguments are really nothing but arguing how many angels one could fit on a pin head.
Paal,
What “—UHI correction done by —-(Jones)” ?
I wish that people would spend a little more time on the SST data set, which is as important as the land temperature. I realize that there are some odd inter-relationships in which land is used to adjust SST, but this seems to me to be more in need of work than UHI.
I quote from communication with GISS
“If you want to investigate personally all 3525 non-rural stations as
well as the 2730 rural stations and decide whether each should be in
that or the other category, that is fine with me. I neither have the
time nor any interest in doing so.”
“GW and UHI are dangerous terms because they are easily misinterpreted –
both describe averages, not universal truths: “global warming” means
“increase of the global mean” not “warming occurs everywhere on the
globe”, UHI means “urbanization contributes in the mean (slightly) to
GW”, not “every urban area contributes to GW”.
I think that when the excrement hits the fan GISS will say they are blameless. It will be the fault of everyone else.
A further quote from communications with GISS
“For some stations
the adjustment is too high, for others too low, but since there is no
systematic bias, in the mean these errors will basically cancel out”
This is not so
Since as GISS describe neighbouring rural stations are used to adjust the slope of urban stations
If any of the neighbouring rural stations are affected by UHI then that will be included in the modified temperature. However if an “urban” station is not affected by UHI the slope will not be modified by neighbouring rural station since its slope would be the same.
Thus they don’t cancel out and UHI would be included in final Global temperature.
Considerations:
1. The highest overall density of measurement stations is in Europe. In North America, stations are dense in the East and on the West Coast. Elsewhere in the world, which tends to be mega urban areas or rural, without the many small to mid sized cities characteristic of Europe and Eastern North America, there is overwhelming bias toward urban locations.
2. In addition to the urban heat island effect, consider also the paving effect. In most of the world, rural areas had no paved areas as recently as 60 years ago. Small towns, farmsteads and road side areas in many rural areas have become paved quite recently.
3. The impact of electrification and later, computerization. Current carrying cables and wires have become ubiquitous in many areas where they did not previously exist. Where there is current there is thermal dissipation.
4. RF excitation of atmospheric H2O. Like a microwave oven on a grand scale.
5. Changes in albedo induced by human activity.
Re #38, Steve Sadlov
Now, that’s a fun thought. Got any numbers ?
My question for GISS is why do they think that any adjustment is necessary for rural station data?
The GISS website shows the precipitous decline in the number of stations since 1970. I would really like to see a graph of the number of rural and urban stations by decade over the past century. After looking through the stations on the GISS site, I suspect that most of the stations which have closed since 1970 were rural.
RE: RF excitation numbers. I do not have any numbers. There have got to be, in the RF communication arena, numbers regarding absorption / path loss. From those, it should be possible to at least make – a proxy 🙂
As for direct measurements of thermal effects, there must be lab work that has been done.
Also, I would imagine the military may have actually tried to measure the impact on air temperature. But perhaps it’s classified?
May be interesting to see what’s out there in terms of studies.
This might be of interest.
“TURN OFF THE HEAT’ IN NEW YORK CITY
re #36: It is interesting to ponder what “Increase in the global mean” actually means. I don’t think that it has anything to do with temperature of the core, or at the Moho, or 10km down. Similarly, I don’t think that it includes the stratosphere, or the troposphere, or that part of the globe that lies above, say, 4000 metres. In the same way, it probably doesn’t include (much the oceans, nor is it likely to include the vast expanses of desert, tundra, icefields etc.
So what does it include? The reality is that the temperature records mostly record the temperature of the flat, relatively low altitude areas, very often close to the ocean, where man lives and makes things, and has built transport infrastructure like airports, roads and highways.
It is pretty evident that even to define a “global mean” is problematic. It is not even clear that it is the “global mean” for the land component. It seems to be a “global mean” of a set of temperature stations around the globe, but oh, by the way, the composition, number and reliability of those temperature stations is changing significantly with time. It is of course possible, as some have pointed out, that this factor alone could account for an apparent rising “global mean”.
It is not hard to conclude that since most weather stations will correlate with population density one way or another, that most weather stations are affected by the impact of man’s activities. That is, the laying of heat absorbing roads and concrete structures, the burning of fuel, use of electricity for airconditioning, lighting, heating etc etc. There are clearly going to be major heating impacts on the local environment. It therefore becomes essential that the temperature records are properly adjusted for UHI effects, and that those adjustments are made in a transparent fashion.
It is interesting to observe the satellite infra-red images at night time and to compare the location of weather stations in relation to urban related hot spots. Warwick Hughes, at http://www.warwickhughes.com/climate/ presents an example in the US saying: “Infra-red NASA satellite image of Illinois, Indiana & Ohio. City heat islands are clearly visible. This heat — from buildings, asphalt, cars etc. — has nothing to do with the so-called greenhouse effect. Yet the IPCC uses temperature records from these cities and many others to give a spurious impression that greenhouse warming is already happening.”
Now, as a lay person, I don’t know whether Mr Hughes is right, or whether the climatologists who supplied the information to IPCC are right, but here is a question that can relatively easily be addressed in a scientific fashion. What is the real position. Are many of the temperature stations correlated with infra-red heat anomalies related to urban development or not? If so, surely they cannot be used to establish a global mean that is supposed to exclude local effects (micro-climates) caused by man’s activity.
As I understand it too, there is a relatively simple way to establish whether UHI effects are in fact present in temperature data series. A 24 hour average may show warming year to year. However, drilling down the next level should be revealing. What is happening to the average day time temperature, and what is happening to the average night time temperature. If both series are parallel, and they show warming along with the 24 hour average, then the station records are probably recording a real rise in temperature at that location. However, the average day time temperature is stable, but the average night time temperature is rising, that indicates that there is UHI effect occurring, and therefore a rising 24 hour average may not be reliable. The reason that average night time temperatures might be rising is because the mass of roads and buildings absorb heat during the day, and release it through the night, thus raising temperatures to a higher level than would otherwise apply.
Going to the heart of the science. If those claiming that the “global mean” temperature is rising want their claims to be taken seriously, then it seems evident that they should release the methods and data that they have used to reach that conclusion (assuming that they haven’t already done so) so that other scientists can check their work, and replicate the conclusions (or not) as the case may be. It surely can’t be acceptable for people from GISS to state “If you want to investigate personally all 3525 non-rural stations as well as the 2730 rural stations and decide whether each should be in that or the other category, that is fine with me. I neither have the time nor any interest in doing so.” Ok fella, but if that is your position, why don’t you withdraw your “science” as not complete. You most certainly cannot claim that you have complied with professional standards of science.
Here’s a different take on UHI. Many towns and cities in New Zealand are critical of the officially reported (on TV) temperatures for their locales, particularly during summer when it seems everyone wants their town to be known as the warmest place for a holiday!
Most official temperatures are recorded at airports and the widely help opinion is that they are considerably lower than the temperatures experienced by the majority of the population.
The image issue is important enough for our local mayor to front a campaign to have the city temperature reported from a central park rather than the airport!
RE: #43. Firstly, to clarify one thing, I would expect both daytime and nighttime temps to be elevated by UHI (more properly, Arthropgenic Thermal Dissipation – ATD). This is because factors such as pavement, structures, heating sources, dissipation from electrical sources, etc, would tend to spike the day time highs as well as elevate the night time lows.
Secondly, so far as I can determine, the process of coming up with a factor or algorithm for “correcting” an array of data for UHI / ATD is non trivial and to my satisfaction, has still not been accomplished. There is still much work to be done.
MTB
You make an interesting point. You may be interested to know that Parker’s article which “proved” UHI did not exist was based exclusively on Tmin (the night time temperature) not Tmax (the daytime temperature). Plotting both Tmin and Tmax shows that the diurnal temperature variation is decreasing as Tmin increases more than Tmax.
Why would Tmin increase more than Tmax?
Rate of cooling lower than rate of warming.
Why can I boil water in 2 minutes, but it takes hours to freeze same said water. Assume starting temprature of 23C.
Oh, yeah. Thanks.