Jet exhaust as climate forcing
Guest Post by Willis Eschenbach
The good folks over at the Berkeley Earth Surface Temperature project have published their paper about urban heat islands. It’s called “Influence of Urban Heating on the Global Temperature Land Average using Rural Sites Identified from MODIS Classifications”, by Wickham et al, hereinafter W2013.
They find no urban heat island effect, saying at the close of the Abstract:
Time series of the Earth’s average land temperature are estimated using the Berkeley Earth methodology applied to the full dataset and the rural subset; the difference of these is consistent with no urban heating effect over the period 1950 to 2010, with a slope of -0.10 ± 0.24/100yr (95% confidence).
Having read the paper, I can’t say I’m surprised that they find no difference.
The W2013 authors have used the MODIS 500 metre dataset to identify urban areas. It’s a dataset that uses the MODIS satellite data to classify the planet into land-cover classes, including urban. The W2013 study uses the MODIS data to divide all the world’s temperature records into “very rural” and “not very rural”. Then they show there is little difference in the trends between the two groups.
Figure 1. Actual MODIS grid overlaid on Henderson Field, Guadalcanal, Solomon Islands. Picture is from 2003, prior to recent development. All gridcells would be classified by MODIS as not built-up, since less than 50% of the area of any given gridcell is built-up area. The ocean is at the upper right. Click to embiggen.
Let me talk a bit about the W2013 study and the MODIS map, and demonstrate by example the reason why I didn’t expect their “very rural” split to have much discriminatory power regarding temperature trends. As an example, let me use one of my favorite airports, Henderson Field on Guadalcanal.
Here is the description from the W2013 study of how they identified the “very rural” stations.
The MOD500 map is available as a raster image, providing a binary classification (urban or not urban) for a global grid with pixels of size 15 arc-seconds.
…
Rather than compare urban sites to non-urban, thereby explicitly estimating UHI effects, we split sites into very-rural and not very-rural. We defined a site as “very-rural” if the MOD500 map showed no urban regions within one tenth of a degree in latitude or longitude of the site. This choice should minimize errors that occur from MODIS classifications in fringe areas. We expect these very-rural sites to be reasonably free from urban heating effects.
This led me to investigate the MODIS “MOD500” map. The MODIS dataset is described in A new map of global urban extent from MODIS satellite data. Basically, it uses multi-spectrum radiation to discriminate between types of land cover, and to identify the signatures of various built-up areas..
Of immediate interest to this discussion is their definition of “urban areas”:
a. Urban areas
In both datasets, urban areas (coded class 13) are defined based on physical attributes: urban areas are places that are dominated by the built environment. The ‘built environment’ includes all non-vegetative, human-constructed elements, such as buildings, roads, runways, etc. (i.e. a mix of human-made surfaces and materials), and ‘dominated’ implies coverage greater than or equal to 50 percent of a given landscape unit (here, the pixel). Pixels that are predominantly vegetated (e.g. a park) are not considered urban, even though in terms of land use, they may function as urban space. Although ‘impervious surface’ is often used to characterize urban areas within the remote sensing literature, we prefer the more direct term ‘built environment’ because of uncertainty and scaling issues surrounding the impervious surface concept. Finally, we also define a minimum mapping unit: urban areas are contiguous patches of built-up land greater than 1 km^2. SOURCE
Now, each “pixel” that they mention above is one of the gridcells shown in Figure 1 above. Those are the actual gridcells used in the MODIS map. If more than half of a gridcell is “built environment” (houses, roads, runways, etc.) then the gridcell is counted as “built-up land”. However, there’s a final hurdle. You need to have five adjacent gridcells of built-up land to have those gridcells classed as part of an urban area in the MOD500 map.
This is kind of an odd definition. It means that any small hamlet is rural, since it won’t cover a square kilometre (about 250 acres). You have to have five contiguous built-up gridcells, which totals just over 1 km^2, for those gridcells to be classified as urban in the MOD500 map.
Figure 2 below shows how much of the land shown in Figure 1 could be built up, covered with houses and roads and parking lots, without anything being called “urban”.
Figure 2. As in Figure 1, overlaid with theoretical possible buildup without any areas being classified as “urban” in the MODIS dataset. Gray areas are 100% built-up, white triangles show 40% built-up. A total of about 70% of the land is built-up. None of it is classified urban because nowhere are there are five contiguous built-up gridcells.
In any case, that’s the history. Figure 3 shows the situation today:
Figure 3. Actual MODIS grid overlaid on current view of Henderson Field. Picture is from January 2013. All gridcells would still be classified by MODIS as not built-up, since less than 50% of the area of any gridcell is built-up area. The gridcells shown in red are the nearest to being classified as built up, as they all have about 40% of their area covered with roads, houses, and runways.
Again, in the MOD500 map, none of these gridcells would be classified as urban. Even if those four gridcells were more than 50% built-up, still none of them would be classified as urban because there aren’t five of them and they aren’t contiguous.
However, the fact that the MOD500 map classifies all of these gridcells as being rural makes no difference—the Stevenson Screen housing the thermometers is still in a horrible location. Here’s why, borrowed from Anthony’s post on the subject. This shows a ground level view of the station taken from the terminal. Note the white box of the Stevenson Screen housing the thermometer and other instruments in the green area just left of center, below the smoke visible in the distance.
Here comes a plane! Weather station visible to the left of the plane’s tail, right of the white mast.
Coming into the terminal…
Hey, park it over here!
Uh, oh, look where the jet exhaust is pointed:
Hmmm, a new high temperature today?
Back to normal.
Now, that’s the problem with the location of the Henderson Field weather station. There are lots of good places to locate the Stevenson Screen in the local area … and it’s in a bad place. This situation is mirrored in many airports and weather stations in general around the planet.
The difficulty I have with the approach of using the MOD500 map to distinguish urban from rural is simply stated:
Many of the siting problems have nothing to do with proximity to an urban area.
Instead, many of them have everything to do with proximity to jet planes, or to air conditioner exhaust, or to the back of a single house in a big field, or to being located over a patch of gravel.
And sadly, even with a map averaged on a 500 metre grid, there’s no way to determine those things.
And that’s why I didn’t expect they would find any difference … because their division into categories has little to do with the actual freedom of the station from human influences on the temperature. Urban vs Rural is not the issue. The real dichotomy is Well Sited vs Poorly Sited.
It is for this reason that I think that the “Urban Heat Island” or UHI is very poorly named. I’ve been agitating for a while to call it the LHI, for the “Local Heat Island”. It’s not essentially urban in nature. It doesn’t matter what’s causing the local heat island, whether it’s shelter from the wind as the trees grow up or proximity to a barbecue pit.
Nor does the local heat island have to be large. A thermometer sitting above a small patch of gravel will show a very different temperature response from one just a short distance away in a grassy field. The local heat island only needs to be big enough to contain the thermometer, one air conditioner exhaust is plenty, as is a jet exhaust …
The only approach that I see that has any hope of success is to painstakingly divide the stations, one by one, based on what is in their viewshed, and exactly how far are they from a variety of ground covers. The work done by Michel Leroy of METEOFrance in 2010 lays out one way to do this, as discussed here.
Because even with the outstanding resolution of the MODIS dataset, it still can’t tell us whether the siting of the weather station is up to snuff, or whether we’re just measuring the temperature of a jet engine at 50 metres … and that’s why I don’t find the results in the W2013 paper at all persuasive.
Best regards,
w.
PS—One curiosity. The published paper is a slightly polished version of their earlier pre-publication paper with the exact same title, available here. The curiosity is the re-ordering of the authors on the two title pages, viz:
Internal politics?
I also note that my friend Steven Mosher is now listed as an author on the paper, my congratulations to him.
[UPDATE] Steven Mosher has pointed out that in the analysis, this particular station (Henderson Field) is classified as urban. I have no problem with that, as it is not a long ways from the capital and is likely classified urban by BEST (although likely not by MODIS) on that basis for their study.
Nor was the classification of that particular station my point, which I quote here from above:
And that’s why I didn’t expect they would find any difference … because their division into categories has little to do with the actual freedom of the station from human influences on the temperature. Urban vs Rural is not the issue. The real dichotomy is Well Sited vs Poorly Sited.
It is for this reason that I think that the “Urban Heat Island” or UHI is very poorly named. I’ve been agitating for a while to call it the LHI, for the “Local Heat Island”. It’s not essentially urban in nature. It doesn’t matter what’s causing the local heat island, whether it’s shelter from the wind as the trees grow up or proximity to a barbecue pit.
Steven also adds the following comment:
My main concern is that people will think the article is about urban/rural (see the title … very rural, not) when the text is about “siting”.
Also, folks may get the idea that airports can never be CRN1 [a measure of station quality] … something which Fall 2011 doesn’t support, as many of CRN1 and CRN2 are at airports.
A clarification might help prevent future misunderstandings and gotchas.
My thanks to Steven for calling that classification to my attention, and also for his clarification.
How does one measure and classify widespread temperature inversions fairly common to Mega – Urbanized areas surrounded by mountains, like LA for example? Is it a metric of Global Climate, or an outlier? Is it Climate or Anthropogenic aberration? I think all temps should be taken in locations and by uniform methods that are Anthropogenically (word?) Sterile . . . then you would be observing the “Net” Temperature, and not a mish-mash of variable-divergent data.
Willis, I notice you are not speaking of how grid cells have changed over time. A grid every year over many years. However, maybe I missed it, haven’t read every word. Is that not the real UHI differences we are speaking of that do affect temperature anomalies, not static view at moments, but over time. Seems Mosher lost track of this ‘time’ as BEST got the best of him.
Other have admitted that urban areas have an offset in the anomalies though those already developed have the same, or close, anomalies. But — the number of developed cells over time have drastically increased as years went by. How can someone not see what these facts lead to… a warmer, not globe, but the readings from a satellite standpoint or the average of temperature stations near development.
If we are speaking of all-city’s-global-averaged-temperature, I have no problem, cities have grown over time and bring with them warmer local temperatures, that is what we do read, but globally, that cannot actually be more than a few percent taking every swatch of this globe into account.
So, I see the problem not as just UHI or LHI but UHΔOT, urban heat differences over time. — that does affect the anomalies over time like in all of the time-series we look at. I see no one addressing this.
Just another phony exercise in modeling. More evidence that climate models shoudl be outlawed as the basis for anything, let alone policy decisions. Willis, good job at exposing this latest fraud.
Its worth pointing out our AGU poster on this last year, which used a more restrictive filter for rural (e.g. modis = 0, airports = false, nightlights < 30, impermeable surface area < 10):
http://wattsupwiththat.com/2011/12/05/the-impact-of-urbanization-on-land-temperature-trends/
I've still got some qualms about all the global UHI work I've seen to-date (including our AGU poster). There just isn't enough station density (or enough rural stations with long records) in some areas to do a good comparison.
The U.S. is relatively easy by contrast, as we have 7000-odd co-op stations, many of which have long continuous records. The U.S. is also blessed with accurate lat/lon coordinates of instruments (at least down to 30 meters or so), something not true for much of the rest of the world. Here is our recent JGR paper on U.S. UHI where we found a rather large signal in the raw data (~14-20% of the century scale trend in minimum temperatures), though it was significantly reduced by homogenization: ftp://ftp.ncdc.noaa.gov/pub/data/ushcn/papers/hausfather-etal2013.pdf
Fig. 7 The frequency of urban patch sizes (x-axis, log scale,
spline curves) for each map (excluding IMPSA and HYDE3).
Note the two peaks for MODIS (orange) and GLC00 (black)
for patches below 2 km2
. These peaks correspond to single and
two-pixel patches, which may indicate speckle in these maps.
These size distributions are unique to remote sensing-based
products and much rarer in the GRUMP (blue) and VMAP0
(red) maps. The modal peak in GRUMP is for patches between
29 and 33 km2
in total area. The majority of these mid-sized
patches are found in East and South Central Asia. The
maximum frequency of the MODIS patches was 5,213 at an
area of 0.67 km2
ed mister jones says:
April 4, 2013 at 1:27 pm
Steven Mosher says: April 4, 2013 at 11:39 Temperature measuring in the middle of an airport has only ONE valid and very important use. It gives the pilots information they need to set safe, efficient take off and landing speeds.
I’ve been flying for . . . too many, 36(?) years – That is new
##################
err you are confusing me with the reader I was responding to
Hmmm, anybody do some transects so there’s some ground truth?
Nice job once again, Willis. Thanks. Love the word ’embiggen’. It’s perfectly cromulent!
Moshpit says:
“i do the work I do and don’t care very much about were the credit is given. acknowledgments or byline? phhht. I do what i do.”
That’s why we love you Steve! You are truly “an honest broker”. Just because we don’t usually agree, doesn’t mean we have any less respect for you. Keep up the good work! (but try to come around to our way of thinking, OK?). Sarc!
It is almost inconceivable that a UHI signal does not exist in the land based thermometer record, and if this signal cannot be found in the data set, then it suggests to me that the data set contains significant problems with resolution.
Is not the best indicator of UHI a comparison between the satelitte data sets and the land based data sets for the period say 1979 to 1996? Of course, the land based thermometer data sets would need trimming to the same latitude coverage as the satelitte. Further, ocean temps would have to be included since these dampen response.
I would have thought that that would represent as good as any starting point for investigation. Then a more detailed study could be conducted using only stations that have the highest siting and are truly rural (this would have to be ascertained by physical inspection) and then compare those selected stations with stations that are situated in urban areas where urbanisation has been growing over a period of say 30 to 50 years.
Once something is fully urbanised, the trend of developing UHI is lost. It is the impact of the growth of urbanisation encroaching on rural or semi rural sited stations that is causing the pollution.
Thanks, Steven. So I take it that you did NOT do a sensitivity test on the MOD500 results reported in this paper, showing the “very rural” stations with and without airports? I’m not claiming it would show anything, just asking if it was done or not.
In any case, the issue is not airports. It is bad siting of the Stevenson Screen, which has nothing to do with airports.
###########################################
Not exactly
The reported tests were done at .1deg
Then, I took the list of stations, and reclassifed all the stations using a finer screen
11km free of modis, plus no ISA 1km ( to address some issues at northern latitudes)
zero nighlights, and no airports.
That classification along with some other sensitivity tests were run. Zeke and I came to the meeting expecting the answer to change…… nope, found nothing.
non results are not that interesting. Engineering wise, of course you blather on about
all the crap you tried. I’v tried anthromes, biomes, nightlights, airports, land cover,
modis urban, modis landcover, 300 meter data, 30 meter data, albedo, emissivity,
population, population density, population growth, etc etc etc. and all combinations.
folks are welcomed to accept the warning that you wont find anything dramatic there, or they can go break their back looking through data for 4 years, or they can say I dont believe you. Or they can realize that the real issue is siting.
there was also a study done of airports versus non airports. nothing to show, except some cooling airports in japan. I believe Muller would have shown those charts to Anthony on his visit, since they were shown to me on my first visit and confirmed the independant work I had done.
In the end, for me, finding something would have much more interesting than finding nothing.
Thats why I liked Zeke paper so much ( and our poster) because there was at least a small signal in the US data.. and a small signal in worldwide raw data, but in the final analysis.. nothing
There are two issues that I think deserve more attention.
A) the small town effect. You will note on my site that this is an ongoing project, but to do it world wide I need to reprocess landsat 28m data. Just time consuming. Maybe its hiding there. Put another way, I can’t rule out that there may be a small effect sub 1km. The hitch is this. Spending a year of data processing to exclude that and not having a publishable result ( hey ma, found nothing again ) isnt on my bucket list anymore. lets put it this way.
I do an airport/non airport study and find nothing. AGW folks dont care. And Skeptics wont believe it. So whats the point? In the begining the point was figuring it out for myself. And it was kinda nice to have some folks say ‘thanks for finding nothing”. As for those who were disappointed when nothing was found.. sorry. if you want help looking for yourself, i was nice enough to build tools for you. have at it.
B) microsite. What is needed is an objective measure of good siting. While leroy is a good qualitative start, it is really lacking in terms of scientific backing.. That is, there are no field tests or measurements to back up the CRN1-5 rating. That’s changing…. But the
biiggest issue here is calculating the ACTUAL viewshed of the sensor. Strangely enough the area directly under the sensor is less important than stuff 100m away. hmm, have a read through design docs for eddy covariance methods for flux towers.
Thanks Willis. I have always had a problem where temperatures are measured (and how) in Aus. This was long before the AGW commotion. There are many things which influence temperature readings. There are difficulties in comparing recorded temperatures 20 years apart let alone 100 or more!
Geoff Sherrington says:
April 4, 2013 at 12:30 pm
Australia has climate records for over 1,000 sites. A few years ago I selected about 50 of those least subject to the hand of man, which I called pristine.
The object was to set a baseline. Ideally, all pristine sites should have a similar trend over time, being natural climate change. This would make a baseline above which UHI would show.
It did not. The complete demolition of methods used for quantification of UHI using regional remote methods follows in these figures.
Sorry BEST, you have no story until you can explain what is going on here.
http://www.geoffstuff.com/Pristine_Summary_1972_to_2006.xls
(Data provided, so please be patient with download time).
#############################
Now THATS more like it. a man who has done some work.
A few questions and I will have a look at it.
1. What methodology did you use for classifying pristine?
a) what quantitative metrics.
b) are these metrics tied to known causes of UHI ( see oke, they havent changed for
decades of research.)
2. What is your data source.
3. How did you process the data.
I will note this. almost every freaky site in the world comes from down under. outside the US it is the only place with long term cooling stations. and urban proxies like population are not very good there ( industry towns report zero population)
Willis bula
From the 2 degrees of separation of the RSYC, thank you.
It perplexes me that in this age, the likes of BEST or other, less diligent, authors feel the need to arrive at a conclusion. What is it and why have these “scientists” forgotten that either no answer or the wrong answer is just as important. Is it really the case that we are in a 24 hour news cycle and that ego, prominence or money mean that say Marcott and co cant bring themselves to say “as a consequence of this study, we are unable to discern an outcome that would allow us to offer tangible insights into this issue ?”
The problem here is the same I see in industry these days: if you can’t dump and troll through datasets, then it ain’t either science or worth it. Manually looking at sites and making a decision is too “subjective” for science these days. Maybe because if you did this, you would have an “opinion”, but if it is just “data”, you are off the hook: hey! that is just what the program says!
It is like looking at the individual proxies of Marcott and saying, each of these has too much variation to tell us solid things about any case, so why would an averaged group do better? GIGO used to be a real concern. Now if it is digital it is gold.
If Darwin were working today, he’d be shouted off the stage for not working with computer models and digital data.
Thanks, Mr. Jobs.
‘Steven, you claimed four gridcells were enough to be counted as an urban area by MODIS in the MOD500 map. That was not true.
A simple “Gosh, I guess I was wrong about the four cells being sufficient in MOD500″ would have sufficed.
Giving me an unwanted lecture about some other method of investigation you advise us to take is just handwaving to distract people from your error. Man up, admit your error, and move on. Then we can discuss your next proposed analysis.”
##########################################
Actually Willis, I havent said whether you are right or wrong because there are two ways
that folks describe what you are talking about. There are over 5000 modis urban pixels that have an area of .67 sqkm. How does that happen?
Fig. 7 The frequency of urban patch sizes (x-axis, log scale,
spline curves) for each map (excluding IMPSA and HYDE3).
Note the two peaks for MODIS (orange) and GLC00 (black)
for patches below 2 km2
. These peaks correspond to single and
two-pixel patches, which may indicate speckle in these maps.
These size distributions are unique to remote sensing-based
products and much rarer in the GRUMP (blue) and VMAP0
(red) maps. The modal peak in GRUMP is for patches between
29 and 33 km2
in total area. The majority of these mid-sized
patches are found in East and South Central Asia. The
maximum frequency of the MODIS patches was 5,213 at an
area of 0.67 km2
And I believe from looking at the actual modis plots that there are areas specified as urban that have areas less than 1km. Rather than call you wrong, I’ll just go look at the data. Solomon islands should be a good choice. and some northern latitude spots. Not a big deal, since one way i cross check modis is by using nightlights, and ISA.. you will find minor differences with all the methods.. meh.
Ps.
Lose the conspiracy ideas about the author order. When I reformated the paper, I put Charollte in first cause she was principle and then just put names down as i remembered them. My bad.
Certainly, I trust Steven Mosher’s diligence in searching for the truth and believe he was surprised to find no significant affect. I do however, think the question of UHI (or Willis’s LHI) is deserving of detailed experimentation. This would involve setting up temporary thermometers say 100, 500 and 1000m going away from the stationary thermometer(and wind direction and strength instruments) in 3 directions at 120 degrees for, say, a week or two. If the results show no meaningful difference at 10 randomly chosen stationary thermometer locations, then the argument would be largely over. We probably do more energetic work on less important issues than this. Come to think of it, experiment seems singularly absent from climate science if we rule out running iterations of climate models…
Willis,
I see you spotted a plane at Honiara International Airport. But it doesn’t seem very crowded.
I know you can fly from there to Nadi and Port Moresby. And there is a weekly Virgin flight to Brisbane. But is that traffic going to seriously affect the temperature record?
Zeke Hausfather says:
April 4, 2013 at 2:46 pm
Zeke, always good to hear from you. Thanks for reminder of the earlier WUWT post on the question. Also, your other paper was fascinating, I hadn’t seen that. I’d recommend it to anyone interested in the question of UHI/LHI. It poses and answers many interesting questions.
Have you looked at the same question(s) using the BEST dataset post-scalpel? It seems like that analysis would be worth doing … so many datasets, so little time …
w.
How much energy did all the Nuclear tests release into the atmosphere back in the 50’s – 70’s?
Willis,
Indeed, too many datasets and too little time. One analysis I wanted to see in that Berkeley paper (but wasn’t included) was the UHI signal (if any) pre-scalpal, as the scalpal may be removing a good chunk of the urbanization.
My project at the moment is to figure out cases in which automated homogenization algorithms fail, either though lots of small breakpoints, lots of trend biases, or sparse station density using synthetic data (similar approach to that of Williams et al 2011). Should be interesting to see when they work and when they don’t once we get the results.
Nick Stokes airports have lots of concrete and lots of buildings even if their stuck in the middle of no where , if you ever act worked on a airport you notice how the ‘pans’ heat up compared to the rest of the place .
The local heat islands, plus the Highest-Lowest sampling are Catastrophic, Anthropogenic, Global Warming.
“…allowed me to reclassify the stations using a screen that used, modis, isa, nightlights, airports, 10km screens, 25km screens. None of them gave different answers…”
S.M., did you try random phone numbers to see if that gave you a different answer?
Has anyone ever correlated quality of science in a paper with number of authors? Is it positive, null, or negative?