Note: clearly satellites can see urban heat, as demonstrated by this recent paper unveiled at the 2010 AGU meeting by NASA. See: Satellites Image the Urban Heat Islands in the Northeast. It can also be demonstrated that the UHI biases the thermometers upwards. As cities grow, so does the increased bias. In that paper NASA says:
The compact city of Providence, R.I., for example, has surface temperatures that are about 12.2 °C (21.9 °F) warmer than the surrounding countryside…
So when you see images like this one above, where the satellites can clearly see the UHI, wouldn’t it make sense to then just look at the biggest low pass filter heat sink on the planet, the oceans, to see what the difference might be? After all, we don’t have urban heat islands in the oceans. Frank Lansner thinks it is worth exploring in this guest post. – Anthony
UAH reveals Urban Heat
Guest post by Frank Lansner
How UAH (University of Alabama, Huntsville) satellite temperature data supports Urban Heat (UHI) as a real and significant factor when estimating global temperatures.
Northern Hemisphere temperatures in recent years:
Fig1. UAH global temperatures trend equals global sea surface temperatures: The black temperature graph – average RSS+UAH satellite NH (Land + Sea) – has a smaller warming trend than the other (brown) land data series – but in fact resembles the cooler Sea Surface Temperature trend. (The blue graph “CSST” is an average of the rather similar SST´s: MOHSST6, HADSST1, HASSST2, ERSST.v3b, HADISST1and Kaplan SST 98.)
The satellite data represents both land and ocean temperatures – and yet they resemble only the SST´s. Why ?
Satellite temperatures and SST do have one thing in common: They are for sure without the UHI warming error from the cities and airports – they are excluding UHI:
Fig2. Now we split the UAH data up in a land fraction and an ocean fraction. Both still seems to yield considerably lower temperature trends than the land data (brown) measured from mostly cities and airports on the ground.
So UAH land temperatures have colder temperature trend than the ground based land temperatures. Are the land-data deviations due to general issues with the satellite data then? Perhaps the satellite data happens to show colder trends for some “known” reasons etc?
Not likely: There is a good resemblance between the UAH ocean temperature trends and then the directly measured ocean data, SST (“CSST”). This shows that satellite data (and thus also satellite land data) are indeed useful and likely to be correct.
So, unless the satellites always starts to fail just when flying over land, the deviation between land data measured on the ground (mostly from cities and airports) vs. satellite land data is likely to originate mostly from the ground based land measurements. This “extra heat trend” seen in the ground based land temperature data may be explained by UHI + possibly faulty adjustments of data and siting problems.
– One more result might also support the correctness of UAH data:
Systems will always seek equilibrium.
On fig 2 we see a pattern of gabs between the UAH land and ocean data. However, after the gabs the UAH land and ocean data these data unite again and thus despite the temporary deviations, they still seem to produce a common trend.
Is it surprising that the temperatures over land and sea will seek equilibrium? Or would it rather be surprising if they did not? What force should maintain a still bigger difference in temperatures between land and see trends?
Fig 3. Lets focus on the temporary gabs between satellite land and ocean temperatures. The green curve represents a de-trended version, just the difference between the land and ocean temperature data from satellite. From fig 3 it appears to some degree that land and sea temperatures align or reaches equilibrium mostly when temperature do not change fast.
Lets take a look at the same phenomenon in the decades just before the satellite age – I use original temperature data published en around 1974-84 for this:
Fig 4. On this illustration we have confirmed, that the land-AIR temperatures are fastest to reach a temperature change “100%”, then the Marine-AIR temperatures comes soon after “80%” and finally the sea water surface temperatures reaches the new temperature level. Again it seems, that after a given time ocean temeperatures and land temperatures tends to find equilibrium. The bac-to-equilibrium-between-land-and-sea-surface-temperatures seems to happen whithin few years, escpecially if general warming/cooling pauses or reverses.
With a reasonable argumentation that also the Land fraction of satellite data is a good indicator of land temperatures, lets look at the “extra heat” seen in the ground based land temperature measurements (mostly from cities and airports). How much “extra heat” do the ground based land data contain?
Fig5. The extra heat in CRUTEM3 land data compared to UAH on NH is 0,103 K per decade.
Fig6. On global scale, the extra heat in CRUTEM3 land data compared to UAH on NH is 0,088 K per decade. (0,23K over 26 years from 1981 to 2007).
If the extra heat in data measured on land is applied to a period 1900-2010 – just to get a rough idea of the possible impact – using 35-40% land area as hadcrut does – we get global extra heat of +0,34 to +0,39 K added to the overall warming of the Earth related to the extra heat occurring when measuring from cities, Airports etc.
0,34-0,39 K is roughly half the supposed global warming 1900 – 2010 , but in this context we cannot claim to have quantitative precision, obviously. But the rough estimate of 0,34-0,39 K suggests that the impact of “extra heat” that cannot be detected by satellites plays an important role when trying to estimate global temperature trends.
The problem of “extra heat” in land temperatures (likely to be UHI and more) is escalated by GISS because they extrapolate the ground based land temperature measurements over the oceans in stead of using real ocean data:
Fig7. In the case of Hadcrut temperature series they use around 35-40% land data when calculating global data, but GISS have a temperature product using roughly twice this fraction for land area as fig 7 shows.
Fig 8 until around 2008 this illustration of land vs ocean temperatures was online at the NASA/GISS website. As we have seen, satellite data indicates that land temperatures from ground has trend around twice the trend of land data from satellite data – and as almost twice the warming trend of SST, ocean data. This tendency is confirmed on fig 8. From 1880 to 2007 we have an ocean warming trend around 0,6K and for land its around 1,2 K – twice.
Again, we saw from 30 years of satellite temperatures that global satellite data matches ocean temperatures rather closely. If valid, then the fig 8 indicates a 0,6 K faulty extra heat, UHI etc from 1880 to 2007.
****
Article from which most graphics where taken:
Review and feedback of the above article by E.M.Smith, Musings from the Chiefio:
“The rewritten past”: http://chiefio.wordpress.com/2010/12/13/the-rewritten-past
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Manfred says: “Actually this issue has already been solved with sound mathematics by McKitrick and supportive of Frank’s results. McKitrick correlated land temperature trends with economic development / land use change etc. (over longer time scales than 1 year). the result was that at least 50% of warming was due to non GHG causes.”
I’m not questioning the the existance of UHI. I’m pointing out that time-series graphs of TLT anomalies and SST anomalies and land surface temperature anomalies cannot be used to conclude that the reason for the differences is solely UHI, and that’s what’s been done in this post. Adding the “Incl. UHI” and “Excl. UHI” to Figure 2 is nonsense. UHI is not the sole reason for the differences.
Frank Lansner says: “Steven and Bob, at least so far I have not found errors in NH numbers…”
Here’s the comparison of Northern Hemisphere UAH Ocean TLT anomalies to Northern Hemisphere SST anomalies for the HADISST, HADSST2, and ERSST.v3b datasets. I’ve used the 5-year smoothing you’ve used in this post. Again, there are significant differences between the TLT data and the HADISST and ERSST.v3b data. Keep in mind the HADSST2 data has the bias from splicing the two source datasets.
http://i53.tinypic.com/2e3oarq.jpg
You wrote, “The “errors” Bob mentioned about the CSST being source of my findings.”
That’s not the only part of the errors in your post that I’ve illustrated for you. You keep overlooking the fact that there are areas around globe where UHI cannot explain the differences between TLT and surface temperatures. In other words, your Figure 2 is implying something that is not supported by the data. You cannot present a graph of TLT, land surface temperature and SST anomalies then assume and imply the differences are caused by UHI.
Bob, its midnight in Denmark and i have a little time to look at your many links etc.
But Bob, is it not a reasonable thing to compare NON-UHI including data from land data with possible UHI including data from ground/cities/airports to evaluate UHI etc?
Please answer this.
The result: If you claim my results are all wrong, I cant see how that should change the general approach I describe. And as a true skeptic, the result is not important, what is important is that we learn the truth about UHI what ever it is, and I think the approach of comparing satellite data with ground data for land can only make us learn more about UHI and other measuring problems. So you and Steven should lighten up, and investigate with curiosity as a driver.
But I will have a look at your links. If your results are true and they shows for example much less UHI than I calculated, then I have learned something about UHI i did not expect.
K.R. Frank
Rephrase:
Bob, is it not a reasonable thing to compare NON-UHI including UAH data over land data with possible UHI including data from ground/cities/airports to evaluate UHI etc?
What ever comes out of this, must be a learning point?
K.R. Frank
Bob… you write to Manfred:
“Adding the “Incl. UHI” and “Excl. UHI” to Figure 2 is nonsense. UHI is not the sole reason for the differences.”
ok, thats a little disappointing Bob. It says more times in the text of the article that the extra heat from ground can be UHI, adjustments and siting problems!!!
Come on! Did you expect me to write all that in the graphic?? I hope the other “errors” are more relevant.
K.R. Frank
Bob this is a nice graph you made thankyou:
http://i53.tinypic.com/2e3oarq.jpg
You compare UAH TLT with some SST´s: Honestly, mostly you confirm that match I would say? that there is a rather good agreement between UAH and SST´s when it comes to the oceans? i can see there is some difference, but certainly not anything that makes my approach irrelevant. I took an averagem the “CSST” to avoid discussing which SST to choose. Is that so awfull?
K.R. Frank
Steve Mosher, you write:
“I’ll suggest that Frank start over and keep it simple.
Start with RSS land and CRU Land. The whole globe, and then do three latitude bands. describe the data and the methods completely.”
Well, theres only ONE reason that you suggest this, and this is because you can D… well see the relevanse in doing UAH – ground based compares 😉 So like it or not, but you reveal that what I did was not that irellevant.
Yours and many other possible ways of attacking this can very well be a good idea. i just presented the main thought.
K.R. Frank
Steve mosher, you write the avreaged CSST graph i used:
“Do you think your average would be impacted by the fact that these datasets are incomplete? ”
Steve I dont have to “think” heres the compare of the CSST with some relevant SST curves:
http://hidethedecline.eu/media/PERPLEX/fig81.jpg
I stated CLEARLY in the article, that my results for now are “obviously qulitative”.
I also have explained Bob in comments, that these graphics comes from an article where i was investigating primarly data up to around 1985, and therefore this CSST was perfectly relevant there. I have RE-USED graphics because they definetely show my point since the CSST just happens to remsemble later SST´s to perfection as you can see. In a context where i only say: UAH Ocean data matches SST´s well, the deviation from CSST to any of the SST you might chose is totaly neglible for the point and the useulness.
I have in no way called this something for peer review at this stage. This IS a qualitative – but good – picture of the situation, and yes to go further it takes more work. I would be very happy to receive a fund to do this to perfection (!) I would love it, dont misunderstand this.
K.R. Frank
On climate changes brought about by urban living.
In tandem with urbanisation in Hong Kong, urban temperature has risen faster than the countryside.
Every time a study looks at a single city and its rural neighbours you see a similar result as above. This 0.05C is utter rubbish.
http://www.weather.gov.hk/publica/reprint/r700.pdf
Athens,
Even through the COOLING period of the 1970`s Athens showed a warming trend in Tmin with a population increase from 1,430,000 to 3,200,000 from 1951-81.
It got warmer when it got colder.
http://journals.ametsoc.org/doi/pdf/10.1175/1520-0450%281985%29024%3C1296%3AIOTUHI%3E2.0.CO%3B2
Steve Mosher, you write:
“I showed you in my earlier graph that there are BIG differences in the trends of UAH Ocean TLT and the HADISST, HADSST2, and ERSST.v3b SST anomalies AFTER 1979. Here it is again:
http://i54.tinypic.com/2edcpix.jpg”
Steve… you present data NOT in 5 yr average as i did. And then because there are more oscillations in the UAH data (yes!) my approach of comparing TRENDS of UAH with SST etc. should be wrong???
Steve, when TREND of UAH-ocean matches SST well, while TREND of UAH-land is considderably colder than ground based land – you just focus on the fact that UAH temps oscillates on very short term more than sea surface temps?
Well, i dont agree. Even though that the type of data on short term has different nature, the TREND is certainly 100% relevant.
This even gets funny: The UAH trend is near the average of the SST´s you have shown.
http://i54.tinypic.com/2edcpix.jpg”
!!
Of course data types are different, but you simply cannot imagine a better match – You should be thrilled by your graph 🙂
K.R. Frank
PS: Steve and Bob: Im not aware of more “errors” from you guys if i have missed something please tell. Ii obviously am fully aware as i said in the article: What i showed so far is “qualitative”. There is room for perfection – especially if the concept is generaly “approved” it would mae sence to go further.
K.R. Frank
Frank Lansner says: “Steve… you present data NOT in 5 yr average as i did. And then because there are more oscillations in the UAH data (yes!) my approach of comparing TRENDS of UAH with SST etc. should be wrong???”
I presented the graph, not Steve Mosher, and I also presented the data with your 5-year smoothing. Please read all of my replies from start to finish on this thread.
You continued, “PS: Steve and Bob: Im not aware of more “errors” from you guys if i have missed something please tell.”
I’ve shown you repeatedly on this thread that surface temperatures trends are higher than TLT anomaly trends in areas of the globe like the Sahara desert where there are no heat islands. Therefore your post is in error. Again, please read all of my replies to you on this thread. It almost appears as though you are purposely avoiding that part of the discussion.
Frank Lansner says: “Steve mosher, you write the avreaged CSST graph i used:
‘Do you think your average would be impacted by the fact that these datasets are incomplete? ‘”
Again, I wrote that reply to you, not Steve.
You continure, “Steve I dont have to “think” heres the compare of the CSST with some relevant SST curves:
http://hidethedecline.eu/media/PERPLEX/fig81.jpg”
You continued, “I stated CLEARLY in the article, that my results for now are “obviously qulitative”. I also have explained Bob in comments, that these graphics comes from an article where i was investigating primarly data up to around 1985, and therefore this CSST was perfectly relevant there.”
I have later in the comments on this thread confirmed that the differences between Ocean TLT and SST are significant. Once again, please read all of my comments to you on this thread.
Frank Lansner replied, “Bob this is a nice graph you made thankyou:
http://i53.tinypic.com/2e3oarq.jpg”
And you continued, “You compare UAH TLT with some SST´s: Honestly, mostly you confirm that match I would say? that there is a rather good agreement between UAH and SST´s when it comes to the oceans? i can see there is some difference, but certainly not anything that makes my approach irrelevant. I took an averagem the “CSST” to avoid discussing which SST to choose. Is that so awfull?”
You miss the point of the graph. The differences are not insignificant.
I’ve had it. That’s it. I’m done playing your games.
Regards.
Manfred: Sorry. Something else occurred to me after I replied to your earlier comment about time lags. This reply:
http://wattsupwiththat.com/2010/12/16/uah-and-uhi/#comment-552615
When we think of time lags in response to an ENSO event, there are different lags for TLT and surface temperature anomalies depending on the part of the globe being examined. Most of the lag between surface temperatures and TLT anomalies occurs in the eastern tropical Pacific (24S-24N, 180-80W). The Surface Temperature anomalies (primarily SST) for this part of the globe lead the TLT anomalies by 2 to 4 months.
http://i55.tinypic.com/2eevry9.jpg
The eastern tropical Pacific represents about 12% of the surface area of the globe that’s presented by RSS for their TLT anomalies (70S-82N). In the next graph, I removed the eastern tropical Pacific data (24S-24N, 180-80W) from the “global” data (70S-82N) by scaling the eastern tropical Pacific data by a factor of 0.12 and then subtracting it from the global data. As illustrated, outside of the eastern tropical Pacific, the TLT anomalies can lead the surface temperature anomalies by about a month.
http://i56.tinypic.com/33u6e80.jpg
The TLT and surface temperature anomalies outside of the eastern tropical Pacific both lag the eastern tropical Pacific, but the two datasets are more in line with one another once the eastern tropical Pacific data has been removed. So I should not have presented the comparison with the TLT anomalies lagged 3 months, unless we were interested in the differences in the eastern tropical Pacific.
Bob, this is your graphic done transparant and laid over mine from the article:
http://hidethedecline.eu/media/UAHUHI/Bobsoverlay.jpg
The black arrow points the the blus SST-average graph i used to show my point.
From this picture it is clear that my blue average FARILY represents the SST.
My approach is to compare land-ground with land UAH, and then ocean-ground with ocean-UAH.
And im sorry, but your graph does not convince me that there is anything wrong with my approach when trying qualitatively to show my point. How can my average blue SST right in the middle of the SST´s you show not qualitatively represent SST ? ? ?
Bob, I have to go through and understand a storm of “errors” from your part before sometimes reaching a real error. Therefore I would appreciate if you where a little patient.
Then it seems that in all these comments i have missed one of your objections against my writing – not intended, you write: “You keep overlooking the fact that there are areas around globe where UHI cannot explain the differences between TLT and surface temperatures”
And I think I found the graph you think of:
http://i43.tinypic.com/if1oh5.png
So here we have a GISS 1200 smoothed temperature area in Sahara where GISS comes up wit more heat that UAH can support and confirm.
I will repeat again-again what I wrote to begin with in the article:
“This “extra heat trend” seen in the ground based land temperature data may be explained by UHI + possibly faulty adjustments of data and siting problems.”
So I am clearly for all readers 100% aware that the warmer ground based temperatures can be much more than UHI.
The most important point of the article is, that there is “extra-heat” in the ground based temperatures which can be revealed by UAH data. In many cases this would be UHI, but i am clearlyaware that more factors plays a role.
K.R. Frank
Frank.
Just for starters you cannot compare the curves by “pinning” 1981 to zero.
Well you can “pin” them but it gives a misleading presentation.
Frank Lansner says:
December 17, 2010 at 3:56 pm (Edit)
Steve Mosher, you write:
“I’ll suggest that Frank start over and keep it simple.
Start with RSS land and CRU Land. The whole globe, and then do three latitude bands. describe the data and the methods completely.”
Well, theres only ONE reason that you suggest this, and this is because you can D… well see the relevanse in doing UAH – ground based compares 😉 So like it or not, but you reveal that what I did was not that irellevant.
########
actually Frank I trying to understand what you think you did.
1. Which UAH Land did you use. Then find out what area of kand that covers.
2. which version of CRU did you use, which area of land does that cover.
3. Do UAH and CRU use the same land mask ( answer no)
4. How did you pick 1981 as the “pinning point”
5. How does one actually “pin” two different series with different error bars
together.
Basically before you toss 5 different series at a methodology it is good to clearly lay the method out with explicit references to the data and good descriptions of the methods. So when Bob Tisdale does something I can always reconstruct what he does.
That allows me to check his work and ask him intelligent questions and he and I can reason together. I think there is maybe one case where I didnt understand what Bob was doing and he quickly corrected my misunderstanding. ( on GISS delting SSTs) That is why Bob and I can talk to each other without acrimony despite our differing views on AGW. I see that he is diligent and careful and clear and discussions are always about the numbers. So, I’m offering you some advice to help you clarify what you did. That’s it.
Ok ,
Frank, to check your work I did the following
Step 1.
First I downloaded the UAH monthly data.
Then I downloaded the monthly crutem. Variance adjusted data.
Step 2. rebaseline the CRU data. Cru measures are on a 1961-1990 basis.
UAH is a 1979-1998 basis. This means that you have to calculate the
jan,feb march etc averages for CRU 1979-1998. These figures are then
subtracted from the CRU figures. Thus, cru and and uah are now comparable
Step 3. Plot them. for the complete record. 1978 dec to oct 2010.
Step 4. Difference them and plot, you get a result with no trend. This means the Land trend for both is roughly equal.
Step 5. take the mean of the difference. its zero.
Basically, I’m pretty sure your nistake stems from a couple of issues.
1. you dont appear to be rebaselining, instead you describe setting all the values to zero in 1981. your proceedure ( both picking that year and the setting to zero) is
odd. You want the baselines to be normalized. And since using all the data is generally advised ( otherwise u might have cherry picking issues) You should use all the data.
Anyway, I’m showing completely different results than you get
Hi Steven!
Please disregard global because as I wrote these data certainly needs more work from my part.
NH:
Heres Bobs data including the UAH ocean and some SST´s they seem to match mine.
http://hidethedecline.eu/media/UAHUHI/Bobsoverlay.jpg
F.R. Frank
Havt to go to dinner + +
I forgot
http://hidethedecline.eu/media/UAHUHI/Bobsoverlay.jpg
Frank Lansner wrote, “The most important point of the article is, that there is “extra-heat” in the ground based temperatures which can be revealed by UAH data. In many cases this would be UHI, but i am clearlyaware that more factors plays a role.”
Wrong.
Frank, your post is titled “UAH reveals Urban Heat”, but your post does not show urban heat island effect. It only shows that there are differences between land surface and sea surface and TLT anomalies. It shows nothing more. The rest of what you wrote is based on your assumptions. You provide no comparisons of urban temperatures to surrounding suburban temperatures, and that is something one would expect from a post that proved the existence of UHI.
You are now saying that all of the “extra-heat” in land surface temperatures could result from “UHI + possibly faulty adjustments of data and siting problems” as you wrote in the post,
-PLUS-
the methods used to infill missing data,
-PLUS-
the deletion of SST data in the Arctic and Southern Oceans, etc.
In other words, in addition to UHI, there are many more things responsible for the differences between TLT and surface temperatures anomalies. But your post starts with “How UAH (University of Alabama, Huntsville) satellite temperature data supports Urban Heat (UHI) as a real and significant factor when estimating global temperatures.” Again, your post does not support that.
Are you aware that the annual variations in monthly land surface data are at least 4 times greater than the TLT data and about 35 times greater than the annual variations in monthly SST data?
http://i51.tinypic.com/34dr6de.jpg
With differences that large, one might expect the trend of the land surface temperature to be higher than the trends of the other datasets.
One last thing, about six months ago, we discussed that GISS no longer uses the dataset you presented in Figure 7. I thought you had finally agreed to that. But you wrote in this post about Figure 7, “The problem of “extra heat” in land temperatures (likely to be UHI and more) is escalated by GISS because they extrapolate the ground based land temperature measurements over the oceans in stead of using real ocean data.” In their LOTI product, GISS does use “real ocean data” for the global oceans where there is no seasonal sea ice. Yet for some reason, you insist on misrepresenting how GISS handles sea surface temperatures.
REPLY: thanks Bob, – Anthony
Steven here’s my email: fel-at-novonordisk.com if you like we can exchange some info.
One thing though, you write: “you don’t appear to be rebaselining, instead you describe setting all the values to zero in 1981. ”
This is not a potential problem because it’s the trends we are comparing. The trends do not change no matter what value you use as start temperature for the respective graphs (as long as you start all at same time of course).
K.R. Frank
How UAH (University of Alabama, Huntsville) satellite temperature data supports Urban Heat (UHI) as a real and significant factor when estimating global temperatures.
Don`t you trust the real data, if these three and the rest of the worlds large urban areas are included in the global mean no wonder it is increasing. urban bias of 0.05C per century utter rubbish.
http://i446.photobucket.com/albums/qq187/bobclive/UHICities.jpg
http://www.lavoisier.com.au/articles/greenhouse-science/climate-change/vincentgray-scam-2008.pdf
http://climateaudit.org/2007/08/17/brazil/
comment, vincent Guerrini
Posted Aug 17, 2007 at 8:56 PM | Permalink | Reply
This should be the beginning of a worlwide/continent by continent analysis of rural v urban sites (unless this has been done). Answer QUIXERAMOBIM. Have a look at Puerto Varas or Punta Arenas, Chile. You’d think that by now ONLY rural data should be used for world surface data analysis. BTW does anybody know why UH satellite data for July 2007 not posted yet? (CRU3 etc has)
good work keep it up.
One thing though, you write: “you don’t appear to be rebaselining, instead you describe setting all the values to zero in 1981. ”
This is not a potential problem because it’s the trends we are comparing. The trends do not change no matter what value you use as start temperature for the respective graphs (as long as you start all at same time of course).
###
I’ll suggest that you start by using monthly data. Its weird that you pick 1981, so just start at the beginning of UAH. Make sure you use CRUTEM variance adjusted
Hi Steven, you Write: “I’ll suggest that you start by using monthly data. Its weird that you pick 1981, so just start at the beginning of UAH. Make sure you use CRUTEM variance adjusted”
Actually i dont start in 1981 🙂
I start with te very first month available from UAH: 1978, dec.
Starting from this month I take data from 60 months foreward (5 years) , take the average and not this averege with centre after 30 months, june 1981. July 1981 is then jan 1979 and 60 months foreward and so on. Its a running 5 year mean, and most important: All data series has been treated in the exact same manner.
So I simply start out using first available UAH data.
K.R. Frank