Correcting for the urban heat island effect in global mean temperature calculations

[bascule]

A rather common complaint I see levied against calculations of the global mean surface temperature is that many temperature monitoring stations are located in urban areas. Urban areas are naturally warmer than the surrounding land because there's all sorts of radiant energy being used by the human inhabitants. However, urban areas comprise a relatively small part of the total earth's surface (including oceans), and a disproportionate amount of temperature monitoring stations exist within these areas.

 

If we continue to follow this line of reasoning, they argue that warming is just occurring within our little urban heat islands and that an undersized sample is clouding our view of the overall mean surface temperature by too many measurements within urban heat islands. It's the urban heat islands that are warming up, they argue, not the entire earth.

 

I'm going to go do a little research and get back to you on this one... but to my knowledge people who compile assessments of the global mean surface temperature (like NASA GISS) are able to make urban heat island corrections against individual stations. How exactly these corrections work I'm not entirely certain of, so I'm going to some research and get back to you.

---

Merged post follows:

Consecutive posts merged

I wasn't able to find information on NASA GISS, but I was able to find out what NOAA does in their calculations:

 

For LST [land surface temperatures] , bias errors may be caused by urbanization over the twentieth century, and uncertainty due to the use of nonstandard thermometer shelters before 1950 (Jones et al. 1990; Parker 1994; Folland et al. 2001). Here we use the LST bias uncertainty estimates of Folland et al. (2001).

 

It looks like the corrections primarily affect the uncertainties, and it seems the farther back in time you go things get increasingly uncertain.

 

So to the best of my knowledge (and again I am a layman) they do not attempt to make temperature adjustments based on station metadata, only adjustments in the uncertainties.

 

I have no knowledge of how the uncertainties for a given station figure into the overall assessment of the global mean surface temperature. But it would at least seem that UHI figures into their uncertainties.

 

As to whether they're doing their maths right, you tell me...

---

Merged post follows:

Consecutive posts merged

Probably the more important point here:

 

The urban heat island effect can, by definition, only affect land surface temperature assessments, since all urban areas are on land.

 

However 71% of the earth is covered with oceans. You don't need to cover the ocean with weather stations to measure it: they can be measured via satellite. It's not as if climate scientists are relying on people running land-based weather stations to figure out the sea surface temperature. This is measured independently and comprises the overwhelming majority of the earth's surface whose temperature is being assessed.

 

http://en.wikipedia.org/wiki/Sea_surface_temperature#Measuring_SST

 

Satellite measurements of SST are far more consistent and, in some cases, accurate than the in situ temperature measurements described above. The satellite measurement is made by sensing the ocean radiation in two or more wavelengths in the infrared part of the electromagnetic spectrum or other parts of the spectrum which can then be empirically related to SST. These wavelengths are chosen because they are,

 

1. within the peak of the blackbody radiation expected from the earth, and

2. able to transmit well through the atmosphere

 

The satellite measured SST provides both a synoptic view of the ocean and a high frequency of repeat views, allowing the examination of basin-wide upper ocean dynamics not possible with ships or buoys.

 

So it would seem the uncertainties are much lower for satellite-based measurements of the oceans. Perhaps climate scientists can use the more reliable ocean data to compensate for the more uncertain data for land? I am not a climate scientist or statistician so I really have no clue how exactly their assessments work.

Edited June 15, 2009 by bascule
Consecutive posts merged.

[the tree]

If the sea temperature can be easily measured by satellite, then why not have everywhere measured by satellite? It would seem sensible to think that satellite measurements are less prone to local errors in general.

[JohnB]

Bascule, no offence taken. I think this is a good idea as normally a thread will jump around to so many different things. It's good to concentrate on a single theme.

 

If you don't mind, I'd like to take a little bit of time to get full links and put my concerns into a coherent pattern.

 

A couple of notes. GISS do indeed take UHI into account. Their adjustment seems to be based on "brightness". Literally looking at a night time sat photo and deciding how bright the lights are nearby. Brightness then gives a ranking which is used to adjust for UHI.

 

And I posted this graph about a year ago and am no closer to finding out why the early temps were adjusted down by so much.

 

Which is more than a bit irritating.

 

Anyhoo, I'll dig through my links and papers and see if I can get you more data on how GMST is calculated and exactly where I have some concerns with it.

 

Fair enough?

[iNow]

GISS do indeed take UHI into account. Their adjustment seems to be based on "brightness". Literally looking at a night time sat photo and deciding how bright the lights are nearby. Brightness then gives a ranking which is used to adjust for UHI.

This brightness to which you refer... Is this in the visual range of the EM spectrum, or the IR?

---

Merged post follows:

Consecutive posts merged

Also, wouldn't you show an average warming trend when focused on UHIs alone? I mean, limit your population sample to these UHIs... take the temperature readings over time, and compare year over year. My guess would be that we'd see a warming trend there as well. Just a random speculation to be added to the mix.

[Mokele]

How dependent are we, really, on places likely to have a UHI effect? Sure, there's instrument records from London and New York, but what about records from Nowheresville, Kansas (pop: 45)?

 

How big and how dense does a city have to be before it generates significant UHI effect? It's easy to point to extremes (as I did above), but what about dense but small cities (Key West comes to mind), or cities which are large but low density (with plenty of trees, yards, etc.)?

[bascule]

If the sea temperature can be easily measured by satellite, then why not have everywhere measured by satellite? It would seem sensible to think that satellite measurements are less prone to local errors in general.

 

The methodology given for assessing sea surface temperature via satellite is:

 

sensing the ocean radiation in two or more wavelengths in the infrared part of the electromagnetic spectrum or other parts of the spectrum which can then be empirically related to SST

 

I don't know if there's a similar technique for assessing land surface temperatures via satellite, but this particular approach seems exclusive to measuring the sea surface temperature.

[JohnB]

This brightness to which you refer... Is this in the visual range of the EM spectrum, or the IR?

I believe it to be visible light. The brighter the light at night, the more civilisation is nearby. I haven't been able to confirm this as surface temps are a large field and I"m just an amateur reading sh*tloads and trying to understand. It's an area I haven't got to yet in detail.

 

Firstly, so there are no misunderstandings. I don't approach this from the POV that the records are "wrong" and I'm not out to prove that in any way. I'm asking the question "Are they right?" Accurate records are vital to the understanding of climate and to the attribution of climate change.

 

bascule has done some work with models and iNow supports the idea of models, and frankly, so do I. However, a model has to be tested against something, that something usually being the temp record. If the record is "out", then the model will be validated against the "wrong" figures and it's projections will be compromised. To me, this would be "ungood".

 

The temp record is also important for the purposes of attribution. Taking the temp rise over a period we attribute certain percentages of that rise to different forcings and feedbacks. (UHI, land use, CO2, etc) Some of these attributions are mathematical (the direct forcing of a CO2 increase) and some are estimated (water vapour amplification).

 

The record becomes important here because small differences in temp become large differences in attribution. For example, if the record shows a rise of .5⁰, we work out the attributions based on this figure. However if the rise was in fact only .4⁰, (a measly .1 degree difference) the effect would be that our attributions are in reality 20% too high.

 

The main reason IMO for concentrating on the US data and data collections is that you lot have the best measuring system in the world, so if you've stuffed up the rest of us are probably in deep doo doo as well.

 

I'm not saying that this has happened, just highlighting the small factors being worked with and the resulting need for accuracy.

 

To the records. (And talking about NOAA specifically)

 

There are two misconceptions;

 

1. NOAA do not use the satellite data in the preparation of their datasets.

Where they get their data for the GHCN dataset is found in the table

http://www.ncdc.noaa.gov/oa/climate/ghcn-monthly/source-table1.html

2. NOAA do not adjust only the uncertaincies, but also adjust the station metadata.

 

NOAA, as bascule quoted use the uncertainties in Folland et al 2001, a paper that I can't find except behind a paywall. Why they don't use the more recent Smith et al 2005 ( http://www.ncdc.noaa.gov/oa/climate/research/Smith-comparison.pdf ) I don't know. Smith finds that his comparison means that;

This more conservative approach produces standard errors that are as much as twice as large as the Jones et al. [1997] and Folland et al. [2001b] estimates for the late 19th and early 20th century. Scales of the low frequency anomaly may actually be larger than those assumed by SR05, but it is difficult to show that for the 19th and early 20th century when sampling is too sparse to estimate them.

 

Note that all this means is that the uncertainty is greater as we go into the past due to poorer equipment and sampling methods. (No sh*t Sherlock. But it's nice to have then quantified.)

 

I have no quibble ATM with the uncertainties, from what I have seen they seem quite reasonable. It's low uncertainties that bother me.

 

Adjustments to the data.

 

This page

http://www.ncdc.noaa.gov/oa/climate/research/ushcn/ushcn.html

lists the adjustments made to the data in a 6 step process. Each step is an individual process and the effect is cumulative.

 

Individual adjustments are shown by this graph from NOAA.

 

This shows the adjustments to data as individual steps, however the combined graph shows the final result.

 

 

So what does the above mean? Blunty, if a temp sensor recorded the same temperature each year (no change at all) then the output dataset would show a .5⁰ F rise in temp from circa 1950 to 2000. Roughly .1⁰ F warming trend per decade, which is around 1/3 of the total warming for the second half of the 20th C.

 

To say the least it seems very odd that the more recent the readings, the more they are adjusted.

 

So that's what happens to the raw data, but are there problems with the raw data?

 

sufacestations.org have been auditing the USHCN for a couple of years now and have found a number of problems.

http://wattsupwiththat.files.wordpress.com/2009/05/surfacestationsreport_spring09.pdf

 

This is the most recent report highlighting the problems. (Including one that I've never seen mentioned before.)

 

The original boxes or Stevenson Screens have in the past been coated with whitewash. That was the standard coating for these items. In 1979, the standard was changed to a semi-gloss latex paint. One could reasonably ask if this effects the readings inside the box.

 

As described in the pdf he used three Stevenson Screens with matched, calibrated thermistors. The results were that the Latex painted screens read generally higher than the whitewash ones.

I found a 0.3º F difference in maximum temperature and a 0.8º F difference in minimum temperature between the whitewashand latex-painted screens.

 

Remember that these differences are against a background of only a 1.2⁰ F temp change over the last century.

 

However, a point that Watts doesn't make but I will. The change of paint at any given station should therefore result in a step change in the record and should be able to be adjusted for. While the change in the aggregate average would be unnoticable and be interpreted as an increase in the warming trend, it should be visible in individual stations. (Must look into this more.)

 

An argument commonly used against surfacestations is that as "amateurs" they are not qualified to assess the various stations. On this I'll happily call BS.

 

Direct from ftp://ftp.ncdc.noaa.gov/pub/data/uscrn/site_info/CRNFY02SiteSelectionTask.pdf we have the siting classifications as used by NOAA, which are used by surfacestations.

 

a) Classification for Temperature/Humidity

 

Class 1: Flat on horizontal ground surrounded by a clear surface with a slope below 1/3 (<19degrees). Grass/low vegetation ground cover <10 cm high. Sensors located at least 100 meters (m) from artificial heating or reflecting surfaces, such as buildings, concrete surfaces, and parking lots. Far from large bodies of water, except if it is representative of the area, and then located at least 100 meters away. No shading when the sun elevation >3 degrees.

 

Class 2: Same as Class 1 with the following differences. Surrounding Vegetation <25 cm.

No artificial heating sources within 30m. No shading for a sun elevation >5 degrees.

 

Class 3 (error 1 C): Same as Class 2, except no artificial heating sources within 10m.

 

Class 4 (error >/= 2 C): Artificial heating sources <10m.

 

Class 5 (error >/= 5 C): Temperature sensor located next to/above an artificial heating source, such a building, roof top, parking lot, or concrete surface.

 

It should be noted that the only "qualifications" that would be required to check classifications are a tape measure and eyes that work. This is not rocket science.

 

The results aren't good. Out of 1,221 staions in the US, 865 (or 70%) have now been surveyed with the following results;

 

Class 1: 3%

Class 2: 8%

Class 3: 20%

Class 4: 58%

Class 5: 11%

 

Which means that by NOAAs own guidelines, 20% of stations are expected to read >1⁰C high, 58% are expected to read >2⁰C high and 11% are expected to read >5⁰C high.

 

So on an impartial basis, 89% of the US stations are reading too high, this doesn't give high confidence for the ROW, does it?

 

To be fair, NOAA has published an answer to the Watts report:

http://www.ncdc.noaa.gov/oa/about/response-v2.pdf

 

I only came across these two reports yesterday so I haven't digested them yet. I do find the NOAA graph comparing 70 Class 1 stations with the entire network very interesting as it shows bugger all difference. So in that respect I'm not sure what to think.

 

As to UHI adjustments for NOAA.

 

NOAA use the adjustments as put forward in Karl et al 1998

http://ams.allenpress.com/archive/1520-0442/1/11/pdf/i1520-0442-1-11-1099.pdf

 

What I really don't understand in Karl et al is Table 7. I can see how UHI would increase the minimum temp at night, but I can't see how it would decrease the maximum temp.???

 

Yet the methodology of pairing rural with nearby urban sites seems sound.

 

There are a few other points and papers concerning the temp records, but this post is long enough as is I think. The above should be enough for discussion and insight.

 

I'm preparing information on the differences between the various datasets that might be illuminating.

[bascule]

NOAA [...] NOAA [...] NOAA [...] NOAA [...] NOAA [...] NOAA [...] NOAA [...] NOAA [...] NOAA [...] NOAA [...] NOAA [...'] NOAA

 

Honestly I don't give a crap about NOAA... I just cited it as an example. NASA GISS is the one to pay attention to.

[JohnB]

Who do you think NASA GISS get their data from?

[bascule]

Who do you think NASA GISS get their data from?

 

Many different independent sources, including data NASA collects themselves with remote sensing.

[SH3RL0CK]

I do think NOAA would be a respectable dataset and should not be lightly disregarded. If this differs from NASA certainly I would question both sets of data, even if NASA data was from multiple sources.

 

As I have stated before, we need to know considerably more about climate than we do. A temperature increase difference of even a few tenths of a degree makes a big difference.

[iNow]

we need to know considerably more about climate than we do. A temperature increase difference of even a few tenths of a degree makes a big difference.

 

I agree that it's always good to learn more and refine our knowledge, but where specifically do you think we are currently lacking?

 

http://www.aip.org/history/climate/timeline.htm

[SH3RL0CK]

I agree that it's always good to learn more and refine our knowledge, but where specifically do you think we are currently lacking?

 

http://www.aip.org/history/climate/timeline.htm

 

Excellent link, thank you.

 

I think the knowledge that is lacking is the following:

 

1) What will be the future temperature rise and the corresponding problems associated with this?

 

2) What should be done to reverse this trend, or at least mitigate it?

 

These are really more economic and policy questions than science. The correct modelling is very important, IMO, because in setting policy the should be some kind of cost/benefit analysis performed. Even a few tenths of a degree rise seems to have big implications regarding sea level rise, rainfall changes, etc. in the computer models. Additionally, we need to be sure the models are correct because if they are wrong, our politicians might spend and regulate (same thing) more than necessary; might spend on proposals that don't work, or worse they might not spend enough to prevent a catastrophe.

 

I'd rather err on the side of caution. But I don't want to throw money away on ineffective solutions either.

[JohnB]

Many different independent sources, including data NASA collects themselves with remote sensing.

Boy are you misinformed.

http://data.giss.nasa.gov/gistemp/sources/gistemp.html

 

GISS Temperature Analysis

=========================

Sources

-------

 

GHCN = Global Historical Climate Network (NOAA)

USHCN = US Historical Climate Network (NOAA)

SCAR = Scientific Committee on Arctic Research

 

I don't see any remote sensing there, do you?

[bascule]

Boy are you misinformed.

http://data.giss.nasa.gov/gistemp/sources/gistemp.html

 

I don't see any remote sensing there, do you?

 

*blink*

 

http://data.giss.nasa.gov/gistemp/2005/

 

Our analysis, summarized in Figure 1 above, uses documented procedures for data over land (1), satellite measurements of sea surface temperature since 1982 (2), and a ship-based analysis for earlier years (3).

 

NASA GISS most certainly makes use of remote sensing...

[JohnB]

*Blink*

 

Well isn't that interesting. One page says they do and one page says they don't. Fascinating.

[bascule]

FWIW according to your URL the remote sensing satellites are also operated by NOAA, not NASA (which I find strange because I maintained the web site of some remote sensing satellites operated by NASA)

 

Maybe NOAA compiles the data... not really sure how that works.

 

Step 4 : Reformat sea surface temperature anomalies

---------------------------------------------------

Sources: http://www.hadobs.org HadISST1: 1870-present

ftp.emc.ncep.noaa.gov cmb/sst/oimonth_v2 Reynolds 11/1981-present

[JohnB]

I was thinking that myself. It does seem odd that NASA wouldn't have their own sats for the GISS.

 

I suspect that GISS get a complete data file from NOAA that contains the sat/sea interpolation as well as the GHCN data.

 

Works been hectic so I haven't had a chance to check out the actual process yet.

 

WRT "remote sensing" sats, maybe the NASA ones you knew had a different sensor? Radar altimetre as opposed to temps? Just a thought.

[Dr Fell]

this is all i can find out about GISS, but it comes via a climate skeptic website, i studied chemistry & don't know enough about atmospherics to comment

 

 

http://wattsupwiththat.com/2008/07/18/cedarville-sausage/