Hansen’s downward adjustments of Peruvian temperature records by as much as 3 deg C is based on the presumptive quality of Peruvian “rural” sites. If one even spends 40 minutes examining the locations of these sites, any resemblance to rural USHCN sites disappears.
In addition, the failure of NOAA and NASA to update their records is notable. In some case, NOAA maintains up-to-the-hour records of sites which GHCN and NASA have not updated in decades. I counted 13 Hansen-rural sites in Peru.
Brief comments on each one follow. Online versions at NOAA are mentioned – see this source citing NOAA (I haven’t sourced the data at NOAA directly at present).
Iquitos: This record is current to 2008. Wikipedia says:
Iquitos is the largest city in the Peruvian rainforest, with a population of around 400000
Yurimaguas: This record ends in 2003. Online at NOAA PR84425. Wikipedia:
Yurimaguas is a thriving port-town[1] in the Loreto Region of northeastern Peruvian Amazonia. Historically associated with Maynas(Pais de los Maynas)[2], the culturally diverse town is affectionately known as the “Pearl of the Huallaga” (“Perla del Huallaga”). Yurimaguas is located at the confluence of rachel the majestic Huallaga and Paranapura Rivers in the steamy rainforests of northeastern Peru,[3]. It is the capital of both Alto Amazonas Province and Yurimaguas District, and had a population estimated at about 64,000 inhabitants (2002).
Moyobamba This record ends in 1988, but is online at NOAA Station Id: PR84435.
Moyobamba is a city and capital of the San Martín Region in northern Peru. There are about 50 thousand inhabitants.
Chachapoyas: This ends in 2001, but is online at NOAA as Station Id: PR84444.
Chachapoyas is a city in northern Peru at an elevation of 2,235 meters (7,657 feet). The city has a population of approximately 20,279 people.
Lambayeque: This record ends in 1961.
Lambayeque is a city in the Lambayeque region in northern Peru. It is notable for its exceptional museums featuring artefacts from local archaeological sites.
Tarapoto This record is current to 2008 and is online at NOAA Station Id: PR84455
Tarapoto known as The Palm Tree City is a thriving commercial city in northern Peru, an hour by plane from Lima, situated in the San Martín Province of the San Martín Region, located to the east in what is known as the selva baja. Although Moyobamba is the capital of the region, Tarapoto is its largest city, and is linked to the Upper Amazon and the historic city of Yurimaguas by now a maintained transandean road. The city is 350 meters above sea level and has a population of over 120,000 inhabitants.
Cajamarca: This ends in 2001.
Cajamarca is located in the northern highlands of Peru, and is the capital of the Cajamarca region. It is approximately 2,700 m (8,900 ft) above sea level and has a population of about 135,000 people.
Juanjui: This record ends in 2001. It is located in northern Peru and is described as a town.
Tingo Maria: This ends in 2001.
Tingo María is the capital of Leoncio Prado Province in the Huánuco Region in central Peru. It has a urban population of around 55,000 (June 2007)
Jauja. This record ends in 1975.
Jauja is a town of 25,000 people in central Peru, capital of a province with a population of 105,000. It is situated in the fertile Mantaro Valley, 45 kilometers to the north of Huancayo (the capital of Junín Region), at an altitude of 3,400 m.
San Juan (de Marcona) This record ends in 1997, but is online at NOAA http://www.climate-charts.com/Locations/p/PR84701.html. It is mentioned as having a population of 11570.
Quince Mil : This is a very high airport in Peru in the Cuzco region. I didn’t locate any population information in a quick look. Record ends in 1985.
El Alto I didn’t locate any population information in a quick look. The record is very short – from 1951 to only 1971.
Hansen’s Population Data
It’s hard to figure out why Hansen would classify (for example) the city of Iquitos (population 400,000) as “rural”. Hansen et al 1999 provided the following definitions for “rural”, “small” and “urban”:
We use the definition of Peterson et al 1997 for these categories: that is, rural areas have a recent population of less than 10,000, small towns between 10,000 and 50,000 and urban areas more than 50,000. These populations refer to approximately 1980.
Peterson et al 1997 is presumably Peterson and Vose (BAMS 1997), An overview of the GHCN temperature database. Here’s how Peterson and Vose 1997 introduced their population data set:
Given the popularity of GHCN, researchers at NCDC, CDIAC, and Arizona State University have prepared an enhanced database to serve the ever increasing demand for these data. This archive, GHCN version 2, breaks considerable new ground in the field of global climate databases. Enhancements include 5) detailed metadata (e.g., population, vegetation, topography) that allow more detailed analyses to be conducted; …
Wherever possible, we used population data from the then-current United Nations Demographic Yearbook (United Nations 1993). Unfortunately, only cities of 100 000 or more inhabitants were listed in the yearbook. For smaller cities we used population data from several recent atlases. Again, although the atlases were recent, we do not know the date of source of the data that went into creating the atlases. Additionally, this represents only one moment in time; an urban station of today may have been on a farm 50 years ago, though it is probably valid to assume that if a station is designated rural now, it was most likely rural 50 years ago. Knowing the importance of avoiding the effect of urban warming by preferring rural stations in climate analysis, these population metadata have been used as one of the criteria in the initial selection of the Global Climate Observing System (GCOS) Surface Network (Peterson et al. 1997a).
These cities have been growing quickly, but it’s hard to believe that atlases in use in 1997 would have classified Iquitos as rural. It’s questionable whether any of these stations are actually “rural” within the proposed definition of Hansen et al 1999. Many of the sites even seem to be “urban” (rather than “small town”) under the definitions of Hansen et al 1999.
If the supposedly “rural” comparanda are actually “urban” or “small” within the Hansen definitions, then the GISS “adjustment” ends up being an almost completely meaningless adjustment of one set of urban values by another set of urban values. No wonder these adjustments seem so random.
Yeah, yeah, Hansen et al 1999 was written 8 years ago, but the same crappy population database is used in adjustments being done as we speak.
And by the way, I didn’t search through the database and pick out Peru as a lurid example. I’ve been to Peru though I haven’t traveled extensively through Peru and it didn’t have too many stations to look at so I could do a quick first cut analysis. So it’s not like I examined Ecuador and it was great and I’m only showing Peru because it’s bad. It just happened to be what I looked at. Maybe Hansen’s done a terrific job on every other country.
Update: This post is criticized by Tamino here where he describes my criticism of NASA’s population data as “Despicable”
Climate Audit 
183 Comments
Steve, you owe me a new keyboard and monitor!
I wonder if some of these stations are considered “rural”, because of the perception of a lack of asphalt, etc (typical of a third world country) that would cause UHI? Certainly, there would be a land use change over the period of record, but how would that change impact the temperature trends? Since each city is different, I would expect a different adjustment for each city. It seems, of course, that true rural stations don’t exist, so adjustments are not possible.
This is interesting …
At first blush it does not seem completely illogical to have a negative value for urban vs jungle temperatures. Of course I my experiences are related to hiking not measuring climate change. In terms of comfort, walking about in the city was much cooler than traipsing about in the forested areas.
However, I have not thought through this in terms of adjustments. Your thoughts?
Even cities of 10,000 can have measureable UHI’s, if the station is located in amongst the dwellings. If it’s a few miles away, it is probably ok.
#2. Irrelevant – “rural” is defined by Hansen in terms of population. He’s not looking behind that. The issue is whether his population metadata is so completely screwed up as to render his calculation meaningless- looks meaningless in Peru.
#3. we’re not really comparing jungle to city, but small cities to other small, medium and large cities. Whether this adjustment means anything looks increasingly dubious.
As an old Peru hand (and CA lurker), I would be cautious about classifying Peruvian towns or met stations. Power supply outside the capital is limited and expensive. This is especially true of rainforest towns supplied with deisel by riverboat (Iquitos, Tarapoto, Pto Maldonado, etc) and the high Andean towns (where the Sendero Luminoso had its roots).
From the mid-80s to early 90s, the country was beset by a vicious terrorist campaign that featured (amongst other stuff) destruction of infrastructure and major negative economic growth.
I don’t have a whole lot of confidence in any Peruvian records, adjusted ot not.
If the GHCN station locations can be trusted (I realize that’s a big if), most of the rural Peruvian sites are close to but not in the towns/cities from which they get their names. It’s reasonable that the sites would be named according to the nearest town/city. That does not imply that they are in the same town/city.
Now that I’ve written this, let me check using Google Earth. I simply did a visual estimate of the distance from the town. GHCN coordinates are given to 0.01deg:
First my summary to save some reading (details below):
Within the resolution of the GHCN coordinates:
– 7 of the 13 stations appear to be rural;
– 3 of the 13 stations appear to be not rural;
– 3 of the 13 stations are difficult to classify;
=====
Details:
Iquitos:
Station seems to be in the middle of the city.
Rural designation is probably wrong.
Yurimaguas:
Stations is ~0.02deg E of town.
Rural designation seems ok.
Moyobamba:
Station seems to be inside the town.
Rural designation is probably wrong.
Chachapoyas:
Stations is ~0.04deg E of town.
Rural designation seems ok.
Lambayeque:
Stations seems to be inside the town.
Rural designation is probably wrong.
Tarapoto:
Stations is ~0.09deg N of town.
Rural designation seems ok.
Cajamarca:
Station is ~0.04deg ENE of town.
Rural designation seems ok.
Juanjui:
Station is ~0.03deg S of town.
Rural designation seems ok.
Tingo Maria:
Station is ~0.15deg NNE of town.
Rural designation seems ok.
Jauja:
The map resolution is poor here.
Station appears to be on the S edge of town.
San Juan (de Marcona):
Station is on the coast on the edge of town.
Quince Mil:
Map resolution is poor.
No communities visible in the area. Nightlights are very dark.
El Alto:
Station is probably at the airport NE of town.
=====
(BTW, the KML file of all stations in the “Googling the Lights Fantastic” thread is an excellent resource. Thanks to Barry Wise, Anthony Watts, and Steve McIntyre for providing it).
I was in Iquitos in 1984. Flew in and out of Iquitos’ airport on a 727. Out was a nonstop to Mexico City. Iquitos is a City. Let me guess where the measuring device is! But I will defer to Jack Linard’s caveat of caution.
Brazil had similar issues.
#7. John V, I think that your calculations here are completely meaningless. You preface by saying:
well, if you can’t trust the population figures, why would you trust the lat/longs other than generically. It’s completely foolish to say that the designation of Tingo Maria, for example, “seems OK”. You have no idea where the station is relative to city limits, nor do you know this about any of the other stations.
OOps, my bad. LOL
#10 Steve McIntyre:
You’re right, I don’t *know* if the stations are inside or outside the towns. Neither do you.
In the case of Tingo Maria, the coordinates put the station ~15km outside town.
A quick google check on “Iquitos weather station” turned up this: http://weather.gladstonefamily.net/site/SPQT
SteveMc, #3,
My mistake. I posted in the wrong thread.
It would be really hard to make the present, or future, hotter now that everyone is watching, but, I never thought the past could be made 3 degrees C colder. That makes for quite a warming trend!
Can someone tell me how you go about adjusting the temperature of a hot, steamy city beside the ocean with data from cities that are in the mountains at 7,000-9,000 ft. elevation? Say there was a warming of 3 C at high elevations; that doesn’t mean one would see a similar warming beside the sea. This whole adjustment nightmare makes me shake my head.
RE12:
We do now.
Funny thing, I used to think the NCDC database was only for US stations…but it actually covers the world.
This is NCDC’s map view using Google Earth API, I see a what looks to be runway and a city near the blue marker, I’m guessing there may be some location error, so perhaps the station is not in the middle of the river as the marker shows.
The popultion data from Peterson et al is taken from a UN yearbook and a variety of atlases. Provenance of teh data is unknown.
One could wonder why, in 1997, they did not use Email to contact the embassey of these countries for the population data taken from aq current census. One could also wonder why they did not use teh Internet to find an Email address for the census bureaus on their own.
Major items of research and policy are based on data taken from atlases. Heaven help us.
RE 7. JohnV one cool thing that no one has done is list all the stations within the adjustment radius.
For example. If a Station is Urban or Small, then Gisstemp will check for stations within
500km or 1000km.
So, it would be a GREAT resource to have that list. For example.
NEWYORK CITY Station xyz: aDJUSTED BY:
A. STATION r1 DISTANCE PDQ
b. sTATION r2 Distance pdq
And so forth. Lots of grief would be saved if people knew which sites where used to adjust which
other sites…
Anyway.. this is a lot more fun than those C02 arguments, stay the heck away from that stuff
I think the point is that while both JohnV and SteveMc don’t know whether these stations are rural or not, neither does, or did Hansen when he used these stations to make these adjustments. To me, that’s inexcuseable.
#17: Anthony, I wouldn’t trust the longitude of that station if I am understanding the metadata correctly. It appears that latitude is given to thousandths of a second accuracy, but longitude has been rounded to the nearest degree.
#17 Anthony Watts:
My goal in #7 was to figure out where the stations really are — this looks like a more accurate data set for doing the same. Good news.
I’d also like to make one more note in relation to this rural vs urban thing. I don’t think it makes hardly any difference. A rural station set beside a tin shed is going to read different than one sitting in the middle of a yard. Which side of a building it is on will also have an effect. From what I’ve seen, pictures, etc, I don’t think any of these thermometers are carefully enough placed to give usefull data, then when you move them? My thermometors at home will read signifgantly different than the thermometer in town the local radio station uses which sits beside a large lake, and it also depends on which way the wind is blowing.
You could be a weatherman.
Well I think it may very well be accurate enough, here is the lat/lon for Tingo Maria, the city:
9°17′43″S 75°59′51″W
The longitude listed at NCDC is simply -76.0, which is pretty darn close. Note that airport is just a bit west of the city center, so while this -76.0 looks like a truncated value, it stands a good chance of being right.
But here is the clincher, if you look up the airport code for Tingo Maria (TGI) you’ll find my hunch was right, they DO list it exactly at -76.0
http://www.world-airport-codes.com/peru/tingo-maria-9307.html
So it appears the temperature sensor is at the airport.
I found a picture of Tingo Maria the city from this page:
Sunset over Tingo María and La Bella Durmiente. Image © Javier Martel
Seems to me like they have electric lights there.
And the airport:
The airport shows that Tingo Maria is not yet a mayor tourist attraction.
Image © Govert-Jan Mennen
Makes you want to visit, doesn’t it?
#30 EW: ~15km => 1.5km-150km. Yes, that’s about the kind of noise I now expect on John V’s signal.
Historic demographic data of peruvian towns (Collated by Jan Lahmeyer)
http://www.populstat.info/Americas/perut.htm
Anthony Watts:
The GHCN coordinates (taken from the KML file from the “Googling the lights fantastic” thread) were off by ~0.17degN and ~0.05degE. If this is typical, it could be a significant problem in manually checking the night lights. Have you seen errors this large in the positions of many other stations?
snip – please stop debating general theory on every thread, folks.
#12. 20. John V – puh-leeze. I didn’t purport to know the exact coordinates of the weather station in these Peruvian cities and whether they are inside or outside of town.
The issue is Hansen’s criteria – sites are defined to be “rural” if the town has a population of under 10,000. According to that definition, Iquitos is not “rural” and Hansen’s classification is rubbish.
If Hansen wanted to define things differently, to have done site surveys and classify locations according to site surveys, that option was open to him. He didn’t do that. He classified the sites according to a horrendously inaccurate population dataset.
Your effort to argue that maybe the weather stations are outside of town – somewhere in the jungle – is just foolish arm-waving and doesn’t rehabilitate Hansen’s screw-up. It’s not my job to prove that the weather station is in the jungle; it’s Hansen’s job to show that the weather station meets his “rural” criterion.
#9 Bob KC
Excellent memory. My favorite in doing that little bit of data collation was discovering
As was generally concluded when Steve was leading the hunt for Waldo last year, the ROW is going to be full of these kinds of surprises.
Think the humidity difference beteween city and jungle might be a factor here?
Anyway, I agree with Steve; inaccurate numbers of people; although inaccurate numbers and half-baked classification schemes and adjustments for stations the adjusters have no idea what they’re like and what they’re adjusting for and how.
Well, it’s hardly surprising, now is it?
I just did the same MMS search for YURIMAGUAS. This datum puts it on the north edge of the town inside of the nightlights area. There is an airport 3/4 of a mile south of this location which is suspicious, since I would think this would be at the airport. Even so it seems well within the environs of the city by both visual and lights criteria
Now if someone could get these lat longs for all the sites out of the MMS in a file with say the Station ID I could update the nightlights file with something better than .01 degree resolution.
More location mystery…
Tarapoto NCDC coordinates: Google Maps view
The San Martin label suggests a different town name?
Google Earth has coordinates for Tarapoto about 12 km to the west.
METAR Information for SPMS (84425) in Yurimaguas, Peru,http://weather.gladstonefamily.net/site/SPMS,
This site provides location information and a google map, plus links to other Peru stations. Is this the station you are interested in?
In John V.’s SWAG of station location relative to population he says ~0.02 degrees east of town could be judged rural….
~0.02 degrees (1.2 minutes) east of town for Yurimaguas is less than 7000 feet from the town. I don’t see how this is rural.
[A minute of arc of longitude on the equator (or latitude anywhere) is one nautical mile. A nautical mile is 6076 feet according to Wikipedia. So multiply your fraction of a degree distance by 60 for longitude differences and by 60*cos (latitude) for latitude differences when looking for nautical miles. To combine them you have to do a bit of spherical trig…but not much]
#30 Steve McIntyre:
I see your point.
The GISTEMP designations are based on population. What is the source of the population data used by GISTEMP to classify sites as rural? My thought was that they did not simply use the census of the nearest city/town, but perhaps had other info that the site was outside the town.
=====
#36 Peter Van Wirt:
I use the rule of thumb that 1 degree of latitude (or 1 degree of longitude near the equator) is ~110km. This comes from the Earth’s circumference of ~40000km divided by 360 degrees. Your value of 7000 feet (2.2km) seems about right. *If* the site was 2.2km from the *edge* of a town I would consider it rural.
Or you could just go to someplace like this
http://jan.ucc.nau.edu/~cvm/latlongdist.html
or this
http://www.nhc.noaa.gov/gccalc.shtml
and punch in your coordinates.
As the context here is Hansen & UHI, the issue here is the magnitude of UHI at a given distance from a UHI centre of a given size. Not “what is ‘rural’?”
A fixed threshold number relative to the “edge” of town, is useless unless you explicitly define how to locate the “edge”. Some UHIs are huge. Others are small. Some extend beyond the “edge” of “town”, some do not. Some weather stations are deep in the UHI. Others are far outside.
From an auditor’s perspective use of an undefined “edge” parameter provides too much wiggle room. The question is, for each station: (1) what is the modern UHI effect locally (this gives you the modern intercept), and (2) what is the slope, going back over time, that needs to be removed (again, locally)?
Hansen postulates that (1) you can correctly estimate the effect from a sample from one part of the globe, and (2) extrapolate it to other parts of the globe. I’d like to see some proof of these assertions before accepting the proposed “adjustment” as is.
It’s 110.94 KM and then: 1st degree (polar) 1.95 km, 1/3 to Equator degree 55.512, 2/3 to Equator degree 96.336, at the Equator 111.312.
That gives our 2×2 grids 29, 157,207 and 222 sq km and the 5×5 74, 392, 517 and 556 btw
Virgin Post here:
I have to give this site credit where credit is due. It seems to me that Hansen et al should bounce thier ideas off this crowd.
I have been following GW (trying to follow the science – I am an engineer, civil) for about three years now.
I have become very disturbed by what I am seeing with all these “adjustments” to critical data.
Is it my understanding that satellite data have been “adjusted” to more reflect these “adjusted” surface data?
If so, then then we need to start over. Raw data cannot be manipulated like this. A clusterf_ck of epic proportions seems to happen when you try to throw around error in a willy nilly way. You can never find your way “back” again.
My $0.02 worth.
Re #43
EJ,
I don’t think it is accurate to say that the satellite data have been adjusted to reflect the surface data. It is accurate to say that there was much crowing by certain AGW advocacy elements a couple years back when an algorithmic error was identified in the University of Alabama Huntsville (UAH) analysis of the satellite data. Correction of this error moved the calculated temperature anomaly closer towards that of the surface data. Looking at a recent posting by Anthony Watts at Watts Up With That? demonstrates that the UAH and RSS satellite data and the HadCRUT index track pretty closely. GISTEMP, oddly enough, seems to be the outlier.
EJ says:
My understanding is the satellites do not measure absolute temperature and must be calibrated against real measurements. More importantly, the satellite data must be adjusted to compensate for all kinds of distortions such as drift. These algorithms are not exact and involve a lot of educated guesses and the people who evaulate these records look at the surface record to determine if the guesses make any sense. In the case of RSS – a climate model is used to calculate the factors used to correct the data. All of this means is that any bias in the surface record would affect the satellite record. However, it is not clear what the magnitude of this bias would be.
Smiles and with all due respect, as T Soprano would impress.
RE: Satellite “adjustments”
There is a difference.
You said “These algorithms are not exact.”
So, a “model” is run on the data…. I am not impressed with that.
Because as you said “it is not clear what the magnitude of this bias would be.”
There is a huge difference between ‘calibration’ error adjustment and RAW DATA MANIPULATION by some mathematical black box (algorithm).
Give me raw data. Let me make my own adjustments after prudent analysis.
I think all raw data should immediately be archived in a RAW place. From the beginning.
My $0.02 worth.
#17 and #21 (Anthony & Carl)
I’m pretty sure that both the longitude and the latitude were originally accurate to the nearest minute (1/60 degree), and then represented as degrees with decimals, which was truncated to five digits after the decimal point.
This truncated decimal degree measure was then calculated back to minutes and seconds. This explains the almost-round 9 degrees 16′ 59.988” of the latitude, which would have been an even 9 degrees 17′ if the decimal reading were 9.2833333333… instead of 9.28333.
I’ve been wanting to ask a question about the validity of UHI adjustments for some time now. To simplify my question, let me discuss a simple scenario of an urban weather station surrounded by 4 equidistant rural stations. My understanding is that the urban station’s trend will be adjusted to match the composite of the 4 surrounding rural stations. I take that to mean the low frequency content of the urban station is what is adjusted.
If this accurately describes the adjustments, then it seems to me that the urban station no longer provides any additional low frequency information content. If the stations are modeled as a common low frequency signal (ie the global trend) superimposed with independent noise, then averaging the four stations will reduce the noise by sqrt(4). If the low frequency content of the urban station is made up of the average trend of the 4 rural stations, then the overall noise reduction remains sqrt(4), not sqrt(5) (which would be the case if the urban site remained independent).
So if my simple scenario is true, how does the urban station help at all?
Steve – I am sure this is off topic and desperate to be moved. Please put wherever it belongs.
If the low frequency content …
… if my simple scenario is true ….
It’s the if’s!
Why guess at something that is knowable? Survey the freakin’ sites and get it over with. Oughta be something somebody could get a grant for or just rely on a surfacestations like approach for the ROW.
RE 37. JohnV the source of Hansens “classification” for being “Rural” is Column 32.
An ascii value that is taken from the USHCN Station_inventory File and GHCN”inv file
There actually was better data available at the time, but he did not to use it.
There is actually better data available from the UN, from Columbia itself, from
the climate studies group at Columbia of which Gavin Schmidt is a member. But he
did not use it. There is actually beter data CITED in Hansen1999, but by H2001 he decided
not to use it. Mystery. there is probably a sensible explanation.
The bottom line is that UHI is not driven by Population primarily. as Columbias climate groups own study of NewYork City shows it is Driven by LANDCOVER mostly. In Gallo98 which Hansen cites, gallo and easterling of NOAA actually studied the use of Nightlights versus using the LULC satiliite products.
Today the serious studies of abating UHI in cities such as ATlanta and Newyork city rely on a
LULC measure and not a nightslights measure. I’ll post the columbia studies on newyork UHI if you like.
Tamino has attempted to claim that the Peru adjustments are legitimate by comparing the site to nearby rural sites. He completely missed the point that removing UHI is justified but removing unexplained urban cooling is not.
He also used two ‘rural’ sites that are not on this list. Does anyone have ideas on why tamino came up with different rural sites? Is this list a complete list or just a selection of sites that steve choose randomly?
RE 50. Pick a Urban site in Peru, Draw a radious of 1000km. Rural sites in that zone will be used
RE 46. PRECISELY, moreoever the urban station is no longer an independenat sample since its value
is dependent on surrounding sites. If you use 10 rural stations to adjust 1 urban station
you dont magically get 11 independent samples. You have 10 independent sample and a frankenstein.
The correct approach is to toss out everything but the best stations selcted according to the
most stringent criteria.
Could the surface station survey be extended to cover all the world weather stations, not just US? I assume both the CA & Surface station websites have readers from all over the world, and could participate in the site surveys?
#49 Steve Mosher:
Ok, so GISTEMP relies on column 32.
It appears that column 32 is not reliable.
Do you know the source of column 32?
I can accept that the rural designation is wrong.
In the spirit of auditing, what is the source of the error? Why does the GHCN inventory file have a rural designation for these sites?
#51, #52 Steven Mosher:
When did you switch to caps on your name?
On the GISTEMP site, you can find neighbouring stations by clicking on the (*) next to a station of interest. This will give a list of all neighbouring sites, sorted by distance.
I agree.
JohnV, Mosher
v2.inventory.inv.txt. The Pop column is used to create column 32. The Pop column and the algorithm to generate column 32 is described in Peterson and Vose section 7.1.
The source of the Pop column is as follows:
– ONC to determine boundaries (Operational Navigational Charts)
– United Nations Demographic Yearbook 1993 (for populations of 100k +)
– “several recent atlases” of unknown source data
#56 conard:
Thanks.
RE 55. I tossed my cookies the other day when I couldnt get at CA so I just retyped
my name using caps. For the record my real name is Moshe R. Steven; But I like
the backwards version.
I know you can click on the asterisk at GISS temp. I was just thinking it would be an
easy little program to whip out. Nobody ever says : here are the 15 Rural stations
that adjust this Urban.
Anyway, stay out of those dang C02 fights: Did you have a look at Lucia’s model?
Lumpy?
On the subject of Peterson–
anyone have a copy of this paper that they would feel comfortable sharing?
I broke a ski yesterday and I am in no mood to pay 30$
Steve: Done
#50. The two sites are in Bolivia and are relatively near Puerto Maldonado.
I said above:
The issue is whether the Hansen “rural” sites are actually rural as represented and what degree of UHI effect may affect the target sites, if they don’t meet the standards that Hansen claimed.
The Peruvian city discussed here, Puerto Maldonado (not Puerto Maldon as named in Tamino’s post) according to an internet source had a population of 28,818 inhabitants.
Tamino’s “rural” comparandum, Cobija, is described as follows:
Cobija does not have a population of under 10,000 (as required to meet Hansen’s definition of rural). So it’s not a “rural” site as Tamino claims and it shouldn’t be used in Hansen’s algorithm either. (In passing, Cobija has almost exactly the same population of Puerto Maldonado. Why should one (Cobija) be used to adjust the other (Puerto Maldonado)? )
It seems that almost all of the sites used to adjust Puerto Maldonado (and other Peruvian urban sites) do not meet Hansen’s “rural ” definition and thus his adjustment is meaningless in this case. That doesn’t mean that the adjustment is without meaning in the US where there is a relatively decent network of rural sites (whatever their warts). Ir doesn’t mean that intercomparisons of stations isn’t useful; but that isn’t what Hansen did.
Boy, Tamino’s a nasty piece of work.
And BTW the most recent data from Cobija only goes to 1989 – so off your Lazyboys, Jim and Reto.
Here’s a function that I wrote to calculate distances. I’ll post up a function to extract stations within 1000 km tomorrow.
RE53, Andy,
Indeed it can be, as the database is setup that way, and I hope it will become multinational. I had to tackle USHCN first, to understand what works and what doesn’t, but even with that understanding, the ROW is a huge challenge for a project like this.
But, it will happen eventually. Its just time and effort.
The debate in this thread is similar to QA arguments at work.
QA: You are not following the written procedure.
MFG: We are doing it the way we think it should be done.
QA: Then you need to change the procedure.
MFG: Then we would have to explain exactly what we are doing, which would be too difficult and/or is improper.
At this point the yelling starts.
#63. I think that the dialogue is more like:
Q: You are not following the approved procedure.
A: You’re an idiot.
re 56
population is not a constant….
Why is it important their nowadays population? I mean, nowadays temperatures suffer no change with these adjustments. The adjustments change mostly the long-past temperature records, so it would be logical to apply a criteria according to wether they were rural or not in the long-past…
I know they said they would use this criteria, and then they are using it wrong. But still I don’t understand why this criteria is used at all.
Also I don’t understand the reason for the adjustment. If anything has happened with any stations in the world, it is that they may have become more urban. More urban means registering higher temperatures. If any adjustments should be made at all, they would be the opposite way, i.e. making the present cooler than the recorded temperatures. Why are they making the past cooler? What explanation do they give?
Regards,
Nylo.
Hansen’s main error is not incorrect classifications of rural or urban places. The main error is assuming that “rural” places have no change in UHI with time. Actually, an Australian study show that places with a small population often have heat islands growing faster than those of large cities. Population numbers are really not very important in determining if sites need corrections for UHI. This should be figured out in some other way, and I suppose the way to do it is to survey the stations, like the people at surfacestations.org are doing.
GISS homogeneity corrections are always continuous, whereas station moves cause discontinuities (jumps). De Bilt and Uccle are classic examples how GISS messes up corrections. IMHO Station move corrections by GISS are far more problematic.
The NCDC website states that the coordinates were originally only posted to the nearest minute. Most of the stations have never been updated. One can see a few at larger cities that have been posted to the nearest second. If the stations are rounded to the nearest minute one is looking at potential errors of 1/2 mile right off the bat. I think you are p***ing in the wind trying to plot the stations. Find the airport. That’s where 95% of small town weather stations will be in South America.
The population logic works like this.
1. if the population in 1980 was rural, it was likely rural before that and so the site
gets no adjustments. even it is urban today
2. If the population was NONrural in 1980, then it is is assumed to have always been been
nonRural and will be adjusted.
This methodology is likely to cause anomalous cases, especially if a city saw large growth
after 1980. its grandfathered in as rural.
But why do rural stations need such an adjustment? Did the people 100 years ago not know how to record temperatures properly? Why does a station that has always been rural need an adjustment?
#70
Or you may get an urban station adjusted (cooled) when the city was smaller and unadjusted, when it is bigger 😉
As to changing procedures, where I work we find that it takes about a full year, and 300 man-hours, to implement a change in a written procedure. We tend to let the “errors” build up until it is worth spending that kind of effort and frustration. I can almost sympathize with Hansen as to not writing down things. Once you write down something, you are bound to it, and can be audited on it. If he has allowed errors to build up, it might be due to the inability of the organization to easily allow changes to be made.
Very few people do any kind of “pure” science anymore. Work has to be tied to cost codes and work packages, time has to be accounted for. We used to have a certain amount of slop built-in to government work, and that allowed for some creativity and productivity. Ironically, with a tightening of control comes a lessening of productivity.
Tables aren’t being updated? Well, they may not have had a line-item in their funding for doing that. Temperature devices aren’t being calibrated? That wasn’t funded this year. It may be next year. If we remember to put it in.
The reality is that there is not a lot of money for basic science, which includes just taking the temperature.
Steve McIntyre:
I’d like to make an attempt at calculating the urban adjustment trends for all stations. (Per Ross McKitrick’s request and some conversation in the previous thread). I think it would be very useful to have a distribution of adjustment trends — perhaps there are patterns from which something could be learned.
As a first step I attempted to download your collated GISS datasets, but the link is broken:
From this page:
http://www.climateaudit.org/?page_id=1686
I followed this link:
http://data.climateaudit.org/data/station/giss/giss.dsete0.tab
(labelled “I’ve posted up my collations of scraped data as follows:”)
I’d appreciate any help finding the collated data. Thanks.
Steve: Go to http://data.climateaudit.org/data/giss/ . I’d recommend using the dset1 and dset 2 versions. I’ve re-scraped the data recently but not reposted.
why not use light data to find urban stations?
oh, you don t like that either..
both approaches wont give you a perfect result, because both are just indicators for the real problem (UHI).
i think the fear mongering (“all the information is false!!!!”) in this sort of post is OVER COMPENSATINg the valuable effect of finding a minor error.
#73
Yes. That is why the post-hoc analysis of estimating and eliminating UHI effects has to be surgical in its correctness and documentation. It is anything but.
#75
The error is “minor”? How did you determine that? Write it up in a paper and publish it. Bye, sod.
You CAN estimate UHI trend effects from a sample and then extrapolate to a population, and then remove the effect from the entire population.
But there are dumb, quick and dirty ways of going about this, and there are smart and correct ways. All we have seen so far are the former.
I certainly did not say: “all the information is false!!!!” and I repeatedly urge readers not to throw up their hands with this sort of attitude.
However, neither is the error here “minor”. Hansen’s analysis is premised on the identification of a subset of “rural” stations from places with populations under 10 000. Seems like a good idea. The problem is that the GHCN population metadata is a farce. In a spot check of Peru “rural” sites, almost none of the sites meet the purported standard – the sites are nearly all “small town” as defined by Hansen, with some actually being “urban” (Iquitos has a population of 400,000).
So what you end up with is the trend not of “rural” sites but of small cities, some of which are growing quite fast.
Let’s suppose that Hansen had reported that the trend in large cities did not differ materially from the trend in small cities. Would anyone interpret that finding as evidence that there was no material UHI effect? OF course not.
And again, I’m not at present taking any position on what proportion UHI may or may not have in measured 20th century temperature increases – which I believe to have taken place. My issue is whether their argument supports their conclusions.
I’m not saying that some other argument isn’t possible; I suspect that it is. As I’ve said many times, I believe that the search for high quality stations with long meta data should be the priority in these studies.
Rural stations do not need adjustment if they truely remain rural.
A town that goes from a population of 1000 to 9,999 may still qualify as rural according to Hansen’s criteria, but the claim that it has had no change in UHI is far from proven.
From what I have seen, such a claim is actually easy to disprove.
LadyGray,
If the alarmists want to use this network and these procedures as the reason for reorganizing the world’s economy and how 8 billion people lead their lives, then they darn well better find the money.
If it’s not important enough to do right, then don’t ask me to change my lifestyle because of it.
bender,
Will you elaborate? I have been working through STEP2 and have been asking myself normalization questions but it seems that the answering self is as vacant as the asking self 😉
Given current giss data {lat,long,elevation,population,nightlight,brightness,etc,}
1. what additional data would need like to have
2. what additional data would you like to have
3. how would you construct an algorithm to test UHI significance for all sites
4. how would you construct an algorithm to adjust for UHI with 95%CI’s
5. how would the adjustments be verified
As far as the GISTEMP adjustments being dumb,quick and dirty I will reserve judgement. FWIW, I see nothing quick about developing a data set like v2.temperature.inv.
I want to see a global UHI trend map, with stations plotted, plus full documentation how UHI trends at station points were calculated and then extrapolated to cover the inner and outer parts of the UHI bubbles. One map for either trend slope or modern day y-intercept (UHI magnitude). One map for date x-intercept (last date when UHI=0). All stations will be “corrected”. The amount of correction will be determined by the UHI map.
I want to see a cross-validation to deternmine how well the global UHI model fits independent samples taken from different regions of the world.
Post the script at CA.
Total cost: one summer student salary. Maybe a Master’s project. Chop chop.
Good point MarkW.
Mark
re: #71 Nylo,
I’ve stated this several times, but the question keeps getting asked so I’ll keep answering. The thing we’re dealing with is the change in temperature over time, not the absolute temperature in given places. This means that one can fix temperatures either in the past or at present or even at some chosen time in between. So there’s no inherent problem with changing past temperatures up or down to make the present temperatures match the thermometer readings.
That having been said, however, I also don’t understand why past temperatures would be decreased to eliminate UHI. If we’ve decided that the average city over 100,000 population has an increased temperature of 1 dec C too high, compared with nearby rural sites, and we want to make present averages, both rural and urban show their present values, then as we go back, temperatures should be raised, relative to the recorded values, so that they show what they would have been had the cities always had their present populations. This would mean that most urban temperatures for the past would be increased, while rural ones would be unchanged.
The way things are done, gives a spurious increase in temperature rise instead of a UHI correction. Either we or Hansen appear to be overlooking something important.
WEATHER STATIONS IN PERU.; Harvard University Gives Up Its System There After Ten Years of Observations.
Special to The New York Times.
February 17, 1901, Wednesday [notice the Y2K date problem]
Page 14, 842 words
BOSTON, Feb. 16. — Announcement has just been made of the abandonment of the system of meteorological stations which have been maintained in connection with the Peruvian station of the Harvard University Observatory. The object for which they were established has been in large part attained, and, with the exception of the station at Arequipa itself, all will be given up.
Full article here: http://query.nytimes.com/gst/abstract.html?res=9405E3D61039E733A25754C1A9649C946097D6CF
#82
My #83 was a crosspost.
You need to estimate the spatiotemporal UHI function. Properly. You need a representative sample of locations around the globe where you’ve installed a high resolution grid of temperature sensors over a year. Enough locations that you have points that are in different states of urbanization along the UHI trajectory. The reason why you need an extensive well-resolved grid is because some weather stations will be deep in the bubble, others – most – somewhere on the fringe. That’s why you need a grid – to span the full range of possibilities. You characterize the whole UHI grid, and that gives you confidence about your one particular point-of-concern within the UHI.
You need to do this across seasons because UHI is stronger in winter at night time than in summer at daytime.
Once you have parameterized the spatial UHI function for sources of given population size, brightness, whatever, in different parts of the world you need to nail down the trend component. This may require a different set of locations as you are heavily constrained to work with those stations that you know have not been moved for a long time – the “top guns”. A “top gun” must have some proxy/index of the source cause of UHI over time. Population mumbers, for example. Or urban canyon depth. I would identify the time-series top guns first, and build spatial UHI grids around those.
Once you have the spatiotemporal UHI function correctly parameterized you are ready to test it against suitable chunks of GISS. AFter that, you are ready to correct GISS.
In the ideal world I would not use GISS to parameterize the global UHI function. It is too sparse to do a good job of it. How good would a GISS-based approximation be? Good question.
Rural will warm relative to Urban.. When RURAL is DEFORESTED over time.
You dont need to poor concrete, or add people, or put up lights. Just
change the land cover from forested. That creates a RURAL heat island
That is why good studies of UHI look at Vegative index, or LULC or impervious
surfaces.
If you want to artifically HEAT a rural location. Cut down the trees.
If you want to cool a urban location. Plant the trees.
The silly thing is that the enviromental studies of Curbing UHI, all indicate
the same thing. You dont need to move people out to curb the heat. you need
to change the landcover.
#74 Steve McIntyre:
Thanks for the updated links.
MarkW (80),
But, if I understood correctly, here we are talking of stations that Hansen considered to be rural. This means that Hansen should not have changed them, but he introduces a slope of up to 3ºC/cent. It could be thought that it is OK if the place was becoming urban. Maybe it did. But Hansen says it is rural. Why does he adjust at all, then?
I have the feeling that they didn’t really look if they were urban or rural at the time of making the adjustments. They said so, but then they told to a computer program: “if you find a spot which doesn’t warm as much as the surrounding spots, change the slope”. They must have used a program to decide which places to warm-up a bit based only on their temperature trends. The result could only be one: there is warming.
I have read from Hansen that the adjustments in the slopes of the temperature trends in “urban stations” meant an increase of the slope in 58% of the sites and a decrease in 42%. Or viceversa, I can’t remember exactly. This would mean that the station being urban or not urban doesn’t change much of its temperature TREND. Of course the temperatures will be higher in urban spots, but will not RISE faster due to climate change. It has little effect on the temperature TREND. The only thing that really affects the trend is when a rural station becomes an urban station somewhere in the time series, or viceversa. So that’s the only thing that needs being identified and adjusted. The rest is crap. An urban station doesn’t need a change in the temperature trend unless you can prove that it was less urban before. And in that case, the changed slope should show more cooling that the raw data recorded by the station.
We will all agree that making a place urban is very unlikely to cause it to lower its temperature recordings, right?
#88 Forest cover can be one of the variables in #87’s list of “whatever”. I suspect that among the world’s weather stations forest cover, brigthness, urban canyon depth, etc. are all fairly well correlated. At some point adding predictors doesn’t get you any furhter – you’ve fully characterized the effect of human population growth over time and space. Additional predictors are largely redundant.
Watt’s compliance index would be another variable to include in the UHI parameterization project.
Without it you do now know how much to correct for non-compliance effects – which are a sub-category of micro-scale UHI effects.
I wouldn’t adjust for UHI. I’d toss out UHI-affected sites and only use confirmed rural, non-microsite-contaminated sites.
#92 typo:
Without WCI you do not know how much to correct for non-compliance effects
RE 85. The UHI adjustments are not designed to cool the past. Sometime they do, othertimes not.
I will give a VERY CRUDE toy explaination.
Take a Site that was Population greater than 10,000 in 1980. This site will get ADJUSTED
Why, because it was NOT RURAL in 1980.
Lets assume that this station showed a 2C Warming over the past 100 years.
NOW, take all the RURAL SITES with 1000km of this site. “average” them. Suppose they Average
1C warming over the past 100 years. This establsihes a 1C warming bias or the urban site.
Adjust the urban site using this 1C bias.
ON the other hand, if the rural Sites within 1000km has shown a 3C warming bias ( from
deforestation for example) then the URBAN site will be adjsuted WARMER to match the rural
trend.
I will say this again. UHI is a function of 3 variables: Surface composition, Urban roughness
( building height) and waste heat.
Population is a poor proxy of this. take a poor proxy and questionable source for that proxy
and you get…..
RE: 90, 91
A paper from Ross McKitrick and Patrick J. Michaels looks at the land use issue.
“A test of corrections for extraneous signals in gridded surface temperature data*”
From the conclusion section:
“Hence, attempts to identify the
magnitude of a global ‘greenhouse’ climate signal on
surface data without properly removing the extraneous
biases (e.g. Tett et al. 1999) risks exaggerating the perceived
influence of atmospheric CO2 levels.”
Paper here: http://www.uoguelph.ca/~rmckitri/research/gdptemp.html
Any quantitative proof of the latter statement? What are the partial correlations in cases where the independent variables mentioned have been measured?
i.e. Is the problem that “population is a poor proxy” or that the population vs time thresholds that are being set in stone are highly arbitrary? e.g.:
10,000 people = rural
1980 = useful cutoff for diserning trend
It seems to me those things are all at least partially, if not wholly, dependent upon population in the first place. I.e., population IS a proper proxy.
Mark
MarkW –
If I were to do a root-cause-analysis of why there are so many errors being found as the data is being sifted and audited, I would point to funding and oversight. Specifically, Hansen et al are not doing the grunt work of collating temperatures and making data tables, that is done by summer hires and work-study graduate students. The work being done by the temporary hires is not being audited, because there is neither time nor money for doing that. Is it possible that some (emphasis on SOME) of the information that is being dug up is possibly either shoddily done or even fictitious? It is possible (emphasis on POSSIBLE). If Hansen tells some peon worker to dig up data on Peru, does he go back and verify that it was done the way he wanted it done? Probably not. We could make some assumptions on what percent of workers are going to be dedicated and truthful in all that they do, guessing that at least 50% are actually doing the work they are told to do. That would leave us with perhaps 50% or less of some of this minor work being wrong or error-ridden.
But I’m not saying that there is that much error. I’m just saying that finding errors such as which towns are rural or urban sounds like someone was fudging their work so they could play more Minesweeper. I don’t think Hansen would do that. I could believe that a college student who was working during the summer might.
Where I work, we have a saying: Trust. But Verify.
Again, the key is not whether the places are rural or urban, but whether they have experienced changes in their “rurality”, “urbanity” or vegetation. Those are the points to address. If they have not been addressed, then no changes should have been made at all.
I would like someone to check what is the trend if we take all of the stations without any of Hansen’s adjustments, only pure raw data. And then to see the average trend of the adjustments themselves.
#100. There are billions of dollars being spent on climate research. If Hansen wanted to allocate a little bit of money to check the population tables, he could have.
As I’ve said before, compare how they do things here to the preparation of the Consumer Price Index. That’s got a lot of mundane detail. But the statistical service is organized to apy attention to the mundane details. All the evidence from GISS indicates that Hansen isn’t much interested in whether the Peruvian data’s been updated (to pick one example.) If the boss of the CPI service were running this operation, he’d be interested in it and have checkpoints on it. Hansen’s the manager of the program and can be held to similar standards. It doesn’t matter whether a summer student in 1995 was playing Minesweeper. Hansen’s responsibility is to manage the program.
He seems more interested in the “big picture”. That’s fair enough. He’s got strong opinions. But if he’s not going to manage the GISS temperature index collection effectively, he should turn over the GISS temperature index to a statistical agency which can manage the operation effectively.
Just a note on “rural” stations. Most of the pictures I’ve seen in “rural” settings, are beside a steel building, or in between runways at an airport, or beside a driveway. 60 years ago, the buildings would have likely been wood, the drives would have been gravel, or, dirt or, grass, maybe brick, but that would be more city streets. Just the changes mentioned there, would change things even in a “rural” setting, independent of any UHI effect. Things change. If you want to “adjust” anything, you are going to have to account for these factors. Good luck with that. JIJO.
It doesn’t matter who actually did the work. Hansen’s name is the one on the paper. So ultimately, he’s responsible.
….
SteveMc,
I know you are opposed to just throwing up our hands, but without accurate metadata, how are we supposed to calculate accurate adjustments for the historical data. If we can’t calculate accurate adjustments, then the data is useless.
When the error bars are an order of magnitude larger than the signal your looking for, you are wasting your time trying to find a signal.
Since we are talking about Peru, let me compare Tingo Maria and Saint-Bruno Qc – towns having roughly the same population.
TM is the centre of coca growing and pasta basica production in Peru. The area has been largely deforested for coca production, electrity generation is limited and irregular. There is one (short) paved road, the runway at the airfield was not paved last time I was there (hydropower studies) about 15 years ago, one flight per week if you’re lucky, virtually no street lighting, immensely poor population, no industry (apart from pb production.
St-B, on the other hand is a prosperous bedroom suburb of Montreal, closer to a forest setting than any of the Peruvian selva towns (TM, Tarapoto, Pto Maldonado) and adjacent to a fair sized provincial park. All roads paved, small industy only.
How could their UHI values possibly be compared on the basis of population? Maybe GDP or PPP indices would be more appropriate.
One size definitely does not fit all.
MarkW –
So Hansen is responsible? Aye, there’s the rub. Who is he responsible to? He has a great job, nice salary, good retirement income to look forward to. He has a good reputation with numerous influential people. He cannot be held responsible, because his managers do not care if he is right or wrong. He has a good public image, which is important for NASA these days.
Steve does a great job with this site, and holds a good moderate line (unusual in these times). He seems to have the respect of a number of people. I thoroughly enjoy coming here and basking in the company of people who understand statistics, probability, computer programming, and basic mathematics. However, unless someone takes the good information that is presented here, and gets it published and acknowledged, it will stay here. A hundred years from now, Hansen will still be known, for better or worse. Will anyone even know of the existence of ClimateAudit?
Being an engineer, I do not mind arguing just for the sake of argument. Will any of this ever be any more than an intellectual exercise in how to not do climate research and analysis? Are there any young climatologists who are reading these posts and getting ready to jump into the fray and disclose their heretical beliefs? Anyone writing papers discussing the way to choose whether a site is urban or rural, and how much to adjust it?
Steven Mosher,
If you want to artifically HEAT a rural location. Cut down the trees.
If you want to cool a urban location. Plant the trees.
Actually, I would dispute this. Relaxing after work with a sundowner, my wife and I are usually treated to an aerial display with the frigate birds gaining altitude for the final leg home to the next island. With the sun setting and shining on the steep hillside where we live, which is covered with dark leaved trees, the thermal is what the birds are using to climb (and amazingly fast, too). So some forests may well cool, but others, probably depending on leaf type and colour, seem to heat. Yes, I know, empirical, unscientific observation, but there it is.
Tony,
It matters whether you are above or below the canopy.
JohnV – the friendliest “keep your enemies close” double agent we’ve ever virtually met.
RE #60
Steve, I pointed out that Tamino’s “rural” sites were anything but on his site. Needless to say I was “snipped”.
As you say he is “a bit of work”.
What’s the status of the Hansen code? Has anyone here got all of the steps up and running?
No it hasn’t been gotten running yet.
#112. It’s still very useful without working. For example the Code 1 flag was identified from reading the code.
Steve McIntyre:
I downloaded the GISS dset1 and dset2 data this afternoon, and just got around to having a preliminary look at it tonight. I didn’t get very far.
Could you point me to a reference for the .tab binary format? (A Google search didn’t turn up anything useful). Thanks.
They are R -objects. Just load them. I’ve posted scripts on this. Here’s an example of a reader http://www.oekologismus.de/?p=726 using this data to produce graphics efficiently (he includes his scripts as well.)
Yes, R, I should’ve known.
This’ll be the excuse I need to finally get acquainted with it.
Thanks.
#117. You’ll thank me for this. It is a tremendous achievement. The turntime for analysis is astounding.
It looks like Puerto Montt would qualify as another one of those unusual sites.
Obviously what’s missing is a baseline, and Hansen’s calcs don’t appear to establish this.
It seems to me that what’s needed is a small sampling of random stations worldwide that
are completely known re history and locations that didn’t change. From there just using
the raw data you could calculate the slope of warming/cooling (if any) and average the
slopes together for the baseline. Doesn’t matter of the sites are at the poles or in a
desert. Just places that didn’t have land use changes.
As Steve Mosher points out land use changes etc would affect the local temp of any given
sensor, whether this is rural or urban. Since the random sampling as per above has already
been thought of (at least here) then this suggests that there are no such sites. I know
we’re cautioned to not throw our hands up, but Mr. Mosher is correct; what is being actually
measured regardless of location seems to be accrued change in land use. Temp is merely the
proxy. Temps OUGHT to go up with population expansion in that case; what was once a forest
is now farmland, etc., meaning UHI is merely a different type of land use change.
Current data suggests that there isn’t enough proper signal to detect the effect of land use,
much less usable enough to differentiate any possible added signal based on CO2.
I also have a question. As I understand it the surface temps aren’t used to calibrate the
satellite data (or are they?) and I’m wondering if so what the relevance is. Why bother with
surface data if sats tell the story? (Besides historically.) Is there a FAQ with a paragraph
or two explaining the basic relationship between surface data and sat data? Finally, what is
the relationship between the sat data and land use expected to be? i.e. if humans are changing
the surface and making locations warmer, doesn’t this also get picked up by sats? Thus the
overall question being this — is the sat data inferring temps ALSO a (secondary) proxy for
changes in land use?
Sorry for the dumb question, but I’m starting to wonder what we’re actually measuring.
Re #88
Indeed, so perhaps when the airport was cut out of the bush at Puerto Maldonado they cleared more of the forest and so the surroundings warmed up and has cooled off since as the trees grew back? If so then Hansen’s methods would have correctly adjusted for the change!
Why would they allow the trees to grow back? Was the airport abandoned?
120 G Alston:
You may want to check out Pielkes’ sites for the literature on land use – they are at least not ignoring the obvious. Second, I do not think it is that easy to size and separate out the different anthropogenic effects that may be applicable to any particular station or sub-set of stations. Third, the real trick is to figure out the extent to which stations accurately represent areas around them w.r.t. all the possible significant anthropogenic influences. Like you, I am hoping that the satellite record will provide a better record going forward, but that leaves many other questions unanswered since as I understand it the satellite readings have to be calibrated against surface readings.
124 Bernie
I think you’ve hit on the problem. It doesn’t really matter whether a station is Urban or Rural. What matters is how land use, construction, relocation, change of equipment, etc, has affected each individual site. Without that knowledge, I don’t think that most increases(or decreases for that matter) in temperature would be outside the margin for error.
There are a group of Monks in N Missouri, that have probably the oldest non-molested weather station west of the Mississippi. I wish I could remember the name. Anyway, their station shows no signifigant warming or cooling. You used to find articles on it once in a while, haven’t seen one in some time. Might be in the GISS maps???
If you can’t filter the noise for each station, how can you use those stations to “correct” other stations????
IMHO, this is worse than the tree ring studies.
re 97 Bender i draw an inference about the effectivenss of Population as a proxy of UHI
from the enviromental studies which show that UHi can be decreased, while population is held
constant, by changing the urban land cover. Planting trees. And also notes that the snows of
kilamanjaro can be decreased by defoestation in the area.
more later
Indeed, so perhaps when the airport was cut out of the bush at Puerto Maldonado they cleared more of the forest and so the surroundings warmed up and has cooled off since as the trees grew back?
How many airports have decreased in size in the last 75 or so years?
re 97 POPULATION and UHI. Gallo99
The populationbased
urban adjustments to the dataset are based on the
analysis of Karl et al. (1988). In development of these
adjustments, Karl et al. (1988) detected the influence of
urbanization on long-term temperature records [the urban
heat island (UHI) bias] for cities with populations
less than 10 000.
No routine adjustments are currently made to global
climatological datasets. The relationship between urban
and rural temperature bias and population has been
found by Oke (1973) to differ between North American
and European cities. Other limitations to the use of population
statistics as an estimator of the UHI bias include
the lack of a globally consistent population data. Additionally,
population data associated with a geographically
defined area can be difficult to relate to the population
(or LULC) in the immediate vicinity of a weather
station. Thus, population alone does not appear to be a
globally applicable method for evaluating and removing
the UHI bias.
Gallo et al. (1993) observed that the satellite-derived
#125
Mosh, let’s not get too hung up on the word “proxy” here. I suggested a multivariate approach to estimating the UHI function, using population as one independent variable of many, including tree cover. Don’t invoke the “proxies are problematic” argument when a proxy approach is not what I’m proposing. I am advocating calibrating the UHI function experimentally, independently of the adjustment process. Which is exactly what you would hope for in paleoclimatic proxy studies – tree response functions calibrated through physiological experimentation.
Re #126
I can think of quite a few that have even closed, including some that I used to fly into. In the case of Pto. Maldonado the logging business is virtually finished, the wild rubber collection (for which the town was founded in 1902) is long gone and more recently so has the Brazil nut collection business, so I think there are plenty of reasons to suppose that economic activity there has substantially decreased.
#129 Phil.
And this supports the principle of using population/development/waste heat as a continuous variable over time to adjust records. Instead, what you have is an arbitrary population and time threshold placed on a development process that is assumed to travel in only one direction through time.
RE: #124 Pofarmer
I found this paper that tells the story of monks manning a weather station in ATCHISON, KANSAS:
mrcc.sws.uiuc.edu/FORTS/histories/KS_Atchison_Doty.pdf
Has photos and a history of the station moves. (ATT: Mr. Watts)
131
Sorry, the link for the Monks in Kansas was missing the http prefix. The hazards of cut-n-paste. Lets see if this one works:
http://mrcc.sws.uiuc.edu/FORTS/histories/KS_Atchison_Doty.pdf
Re #130
Agreed, so we’re also agreed that the procedure that Hansen uses is not the one you characterize above?
Hey the Mo monks. Conception abbey
re #134
A sad note regarding the long time weather observer at the abbey who gained the nickname “the Weather Monk”:
– Cooperative Observer Killed in Shooting Incident
Brother Damian Larson, longtime Coop Observer at Conception Abbey in Northwest Missouri, was killed by a gunman on June 10, 2002. A 71-year-old man walked into Conception Abbey with an assault rifle and murdered Brother Larson and Reverend Philip Schuster and wounded two other Monks before killing himself.
Larson, 64, was from Wichita, KS, and worked as a groundskeeper at the abbey, about 30 miles north of St. Joseph, MO. He was known in Missouri as the “Weather Monk.” His weather cartoons appeared in several small northwest Missouri newspapers.
“Brother Larson continued the tradition of fine Coop Observers at Conception Abbey and served as a Coop Observer since 1969,” said Bob Bonack of Central Region Headquarters.
The Coop Station at the abbey is one of the oldest in the Central Region with records dating back to 1883. Conception Abbey was founded in 1873 by two Benedictine monks from the ancient Engelberg Abbey in Switzerland.
One of Larson’s predecessors, Father Adelhelm Hess, served as an Observer from 1894 until 1964 and was honored with the Jefferson Award in 1963.
From: http://www.nws.noaa.gov/com/nwsfocus/print/printfs061902.htm
These two stations, in Missouri and Kansas, are very close to each other. A direct comparison between them may be edifying.
=====================
Steve McIntyre and Anthony Watts are indeed doing a great job at auditing the data and how the data is used. Obviously good, reliable data exist and sound, honest analysis can be attempted. Now the question alluded to by LadyGray post 107, what can be done so the audit becomes a reference known to others? Publish in a science journal or do we have to go to a Court of Law -and which one- and build a case in order to set the record straight?
#107, 137. I realize that there are a number of things discussed at this blog that are worth writing up for publication in a journal.
Steve –
There is a long-standing tradition in science, going back at least to Newton, where the most prestigious or popular scientists use their influence to squelch dissension and prevent publication of legitimate alternate views. This was seen during the time that Einstein was popular, when several important physics principles had to be postponed until his death, thus setting physics back several decades. Some of the scientists I work with have voiced their concerns over similar incidents in their fields of study, thus indicating this to be a current phenomenon in many fields. Do you believe this could be a problem for publishing at this time?
re 128. Yes, I realize I did …that as I probed into this matter more. OKE 73 ( seen it cited, have not found it online ) “establishes” a Log(pop) relationship for UHI. Subsequent to that he has studies showing that this log(pop) varies by continent. Most of the multivariate work I’ve seen ( super fast scaning of abstracts) has been focused on characterizing the Variability of UHI within the city.. Even A nice little use of PCA..
There is some fascinating stuff out there on UHI. I leave it at this. Population is blunt
axe. fair enough. Hansen had better tools at his disposal. Now, going forward here is the question.
12 years from now… will GISS still use 1980 population figures to define Stations in the ROW
that DONT het adjusted? In 2020 will 40 year population data be used
We could just assume that this disparate mixture of methods for adjusting temperatures was not done out of malice, but occurred because it was cobbled together over a span of several decades. In that case, what if we would brainstorm a sensible and well-documented set of standards for how to adjust temperature? That could be the basis for publishing a paper, and could be the basis for a new international standard (brainstorming means I don’t have to be rational), similar to IEEE standards.
Some examples could be: How to determine the urban or rural nature of the area around the monitoring station, how to determine the area through which the monitoring station is valid, how to assess the amount by which neighboring stations can affect another station, how to determine how elevation or terrain can effect the monitoring station, how to account for areas that have no monitoring stations.
Just the effort of writing such things down can help to expose some of the problems. But, by writing them down in a coherent way, it is possible to come up with methods of overcoming the problems. As people have mentioned, there is very little consistency to the methods, how they are applied, or how often they are upgraded. When I was younger, I hated the term Formality of Operations. I have to now acknowledge that there are definite times and places for formality of operations.
Steven Mosher said: (February 27th, 2008 at 11:50 am)
When I emailed GISS management about why the reference period of 51-80 was used for the GISS anomaly chart, the short answer was “We use it because we’ve always used it.” I assume they’d have pretty much the same reply here…
The abbey at Atchison Kansas. Would that be considered Urban or Rural?
re: #143 pofarmer- Atchison Kansas
“In June 1991, the instruments were moved to a location within the courtyard of the Abbey. The observer requested that this move be made due the vandalism that had been taking place at the previous location. The courtyard is 125 feet square, surrounded on all four sides by 5-6 story brick walls. See Figure 6 for a picture of the courtyard taken in June 2004. The actual move was 200 feet from the previous location. A MMTS thermometer system continued to be used.” see link: http://mrcc.sws.uiuc.edu/FORTS/histories/KS_Atchison_Doty.pdf
By the way – who clicked their heels three times? Weren’t we in Peru? 😉
These posts are criticized by Tamino here where he describes my criticism of NASA’s population data as “Despicable”
I don’t know why I bother ever visiting Tamino’s site. The man has no intellectual credibility whatsoever. After completely mischaracterizing Steve’s argument, he then complains that because Steve doesn’t accept at face value the “adjustments” that Hansen has made, that Steve can’t be trusted to auditing.
Once you convince yourself that data collecting stopped after 1988, it’s easy to believe that populations stopped changing in 1980.
re 145
Can you imagine what it must be like to be a young researcher in that field? The pressure to conform must be enormous
You are famous again, Stan. PZ Myers rant about you on Pharyngula last fall was linked on Tamino’s latest post, and was excoriated.
================================================
Dear folks, Tamino…Kim, I appreciate your brave efforts
over there but 99% PBS?? (Pearls Before Swine) In reality
there are NO, 0 truly rural climate stations, outback arrayed
networks on different altitudes, up-wind, down-wind whatever
is causing warming/cooling, since the treemometers
accuracy is not what you’d wish. BTW Tamino is just a Mozart
opera figure in the “Magic Flute” I Wikied “Tamino” and
what a coincidence: The first one to play “Tamino” was a certain
Benedikt Schack (Schack=chess in Swedish) Referring to a recent
post of mine on another Peruvian thread? if black=warm or warming
and white=cool or cooling Earth is a Zebra or a chess-board…
And Benedikt means “wellsayer”…hmmm….
We could hope for such a global outback climate network
but in 500 years how much space is left…Sorry just got
an Ehrlich-Borgstroem virus…
re 148. Great pressure makes diamonds.
Thou hast transgressed!
I went over there and looked. Cripes, it’s NASA, not the Vatican, last time I checked NASA was wrong plenty of times, they aren’t infallible by a long shot (as two complete shuttle crews families will sadly attest to).
Re #145 Steve:
Tamino seems to have graphs past 1988. Where did that come from? They do not look like the charts in your other post.
Nor am I a regular poster here. Just comparing the CA argument to Tamino’s counter-argument on this topic as an outside observer. CA is winning hearts and minds hands down. All of you at CA should be congratulated for keeping the mudslinging noise level down. I don’t agree with LadyGray that what goes on here will soon be forgotten or make no lasting contribution to this great debate of our day. Of course, peer reviewed publishing is useful. However, this debate goes well beyond the climate science technocrats (whom, if Tamino is any measure, have already settled the science to their satisfaction) to the policy makers, journalists, social scientists and right down to individual lay person voters. This blog is extremely important in reaching out and informing the larger community, especially as an alternative POV in a world of incurious consensus and boilerplate media sound bites.
Steve deserves kudos for moderating fairly and transparently while tolerating dissent. The gestalt that Steve and the regular posters imbue this site with is one of honest, if skeptical, questioning of status quo assumptions.There is a sense of sincere curiosity and good humor here that is lacking at RC, Tamino and Rabbet.
In comparison, Tamino’s S America temp adjust argument is shrill and acrimonious. Even if Tamino’s logical argument wasn’t troubled, his hostile style is so brutal that it’s depressing to read the length of his posts to replies. It’s ironic that the DIY can-do spirit of inquiry that dominates CA is labeled Denialist by Tamino whose argument consists mostly of denying possibility after possibility, while refusing to check even basic facts with a simple Google search. CA is all questions. Tamino never doubts a thing, everything is known, settled and didactic. He is entirely lacking in curiosity and impatient with seemingly foolish questions, not traits one might expect in a natural scientist.
Why Tamino is so angry and defensive? Why does he hide his identity? He seems frightened of something. Why does he suppress all dissent, but those he selects to publicly humiliate? And finally, as Stan points out, what a magnificently anti-Enlightenment ethos to drop an impressionable young graduate student into.
This is what one with no dog in the hunt sees.
I agree with No Dog in Hunt. I also checked Tamino’s place and found it very unpleasant. Very agresive comments and responses.
I also agree that, published or not, all the details being discovered here are of a high importance. Because all of the downplaying of this informations that is currently taking place will start to change as soon as the temperatures start to drop and ice cover starts to increase. Which is likely to happen anytime soon. After all the warming trends have stopped for the last decade.
snip
Re Tamino:
How can someone who critiques Steve and CA for not publishing keep his own papers secret?
#154
NDIH
If you have sparred with a fully committed environmentalist, you will recognize that Tamino’s behavior is not exceptional. Left or right, fanatics are fanatics. As such paranoia, hyperbole, invective and ad hominems are standard. Some regulars here display the same patterns of behavior. Steve McI, to his credit, keeps his own cutting comments to a minimum.
Bruce Foutch says:
February 27th, 2008 at 8:58 am
Nice find. I wonder how many more reports like this are out there?
Re # 159
Take the link back a step, and a lot more histories show up:
mrcc.sws.uiuc.edu/FORTS/histories
Duh!, (/hits head) Thanks, I didn’t get past seeing “FORTS” and ignoring poking around there.
LadyGray says:
These things have been done many times, the last work on most of these things being that Czech PhD Thesis I mentioned elsewhere (methods of missing or erroneous data homogenization, if that should or shouldn’t be done, studying correlations between neighboring stations, influence of elevation differences and season on correlation between stations, etc…).
The problem is, that similar studies are mostly regional, also not in English. To get the things right, the authors have to actually get familiar with the stations, analyze the history and metadata, search in the libraries and archives.
There was no citation of a global study doing similar things, however.
Someone at Tamino seems to have a problem with my comment above:
This was my response based on my current understanding of the issue:
The adjustments for removing spurious *heating* caused by urbanization should result in reduced trends in the urban records. Spurious cooling caused by urbanization has not been identified as a problem which means one cannot justify ‘correcting’ it.
If an algorithm to remove UHI effects results in urban sites having *heating* added then that is good sign that there is a big problem with the algorithm. In this case, the problem is likely a result of ‘rural’ sites that are not really rural.
The after the fact justifications by Tamino (i.e. the site is obviously bad and should be adjusted) make no sense since removing UHI effects is the stated purpose of the algorithm.
The problem is the researcher looks to be trying to fit the data to the storyline——–again.
Unfortunately that is characteristic of those with a certain political bent.
This is what happens when you throw data willy nilly into the alogorithmic meat grinder.
Why would a RURAL site show a greater warming trend than Urban site? Simple. UHI is a Log(pop) phenomena. A rural town of 5000 that adds 5000 people will show a stronger trend, than a city of
1Million that adds 5000. You get backwards adjustments ( rural sites adding warmth to the URBAN sites) when the UHI is saturated at the Urban site but increasing
at the RURAL site. Remember ITS THE TREND at the rural that adjusts the Urban. If the Rural warms less than the urban, then the urban is cooled. If the rural warms MORE than the Urban, then the
Urban is warmed.
Lets take a simple example. Vegas is UHI infected. Adding 10000 more people will not Impact
its temperature trend. ESPECIALLY if OKE73 is correct UHI bias = .73Log10(POP).
Hansens algorithm will WARM a urban center as its rural neighbors are built up.
The TREND of urban and rural are forced into agreement.
a rural location that moves from a population of 1K to 11K will see a big trend, If OKE73 is right and UHI = .73Log10(pop)
That trend, the rural trend is used to adjust the urban trend. So if the urban center
is thresholded and its surounding rural areas are growing, the hansen adjust will WARM
the urban center so that it matches the rural.
There is a better way
As Mosh says, applying an algorithm to an entire dataset which is as poor as the world surface temperature record is a very hazardous project. I would be extremely careful of finding some logical (or illogical) algorithm, testing it on a few cases, and then applying it across the board. Do that, and you end up with this kind of foolishness. You may well be better off taking the time to do it individually.
However, if you are hell-bent on doing it all at once, there’s a way to do it right.
1. Go through all of your data with a variety of means (graphs, derivatives, statistical analyses, violin plots, whatever you can think of) to identify the individual station datasets that have problems. Deal with these as appropriate (toss or fix bad data points, eliminate or fix bad stations, verify the data with the originators, etc). Do this more than once, and more than twice. Use different methods, apply different tests. Make sure the data is good before proceeding.
2. Apply your hot new whiz-bang algorithm to the data.
3. Subtract the original data from the “corrected” data, to create a new dataset. I’ll call it the “subtracted” dataset.
4. Go through all of this new subtracted dataset with a variety of means (graphs, derivatives, statistical analyses, violin plots, whatever you can think of) to identify those individual datasets that have problems.
5. Figure out exactly why and where the hot new whiz-bang algorithm is going off the rails.
6. Change the hot new whiz-bang algorithm so it doesn’t go off the rails in those spots.
7. Lather, rinse, and repeat until step 4 finds no problems.
8. Do global, regional, zonal, and local averages using the original data and the “corrected” data. Make sure the hot new whiz-bang algorithm is not leaving Waldo somewhere unknown …
Although I certainly can’t prove it, I would strongly suspect that the good folks looking after the surface station data have scrimped on some of these steps, or eliminated them entirely. But considering that they lost Cobija for twenty years, there’s no telling what they may not have done.
w.
re 166. I agree with Willis Eschenbach about how one should proceed in data analysis as that is very close to what I do when creating a new software product. I am very careful on checking data on input, intermediate and output steps and I look for anomalies and patterns of suspect data. I got very good at sniffing out programming bugs in maintaining existing software by doing what i call pattern recognition. Seeing a type of suspect record once and it could be an outlier but if you see it multiple times it usually is a programming problem.
re 166:
The worldwide algorithm has one drawback: the statistics are not identical per lattitude. E.g. The station-to-station correlation between near-equator annual temperatures is extremely poor. The reason: The equator region has no winter. Adjusting one equator station using another is therefore meaningless.
re:165
Steve hits the nail on the head by showing how the algorithm mostly tends to warm the really big cities instead of cooling their growing and possibly more UIH affected rural neighbours.
There are ways to make the algorithm better to avoid this. Instead of adjusting a big city according to the average trend of its rural neighbours, you should only do that if you CANNOT find a rural neighbour with a similar trend to that of the city. That is, adjust with more heating or more cooling only if ALL of the rural neighbours show a higher warming / cooling. The problem can still exist but is more unlikely to happen. It’s dubious that all of the rural neighbours will be growing in population. I would like to see this implemented…
LadyGray, ca: 100 – “Trust, but verify.”
You work in Russia?
MarkW says on February 27th, 2008 at 4:00 pm:
Cheap entertainment? I visit maybe twice a year, but don’t stay long – I suspect that snickering & sneezing on my keyboard isn’t good for it. Comments from the choir always stick to the same theme: The enlightenment of the climatologists, and the mean-ness & pettyness of Steve – always trying to derail their noble mission to Save the Planet. Even the name – I always think of Bertrand Russell’s caustic observation that an open mind is an empty mind.
Mosher 165: Nice clear explanation by example, so that even a sideliner like myself got it. That, and the sound reasoning supporting it – exactly what’s missing at Empty, er, Open Mind.
Hans Erren:
Here are some nice correlation pics of the Czech stations influenced by season and differences in elevation and distance.
Here’s graph for winter and summer:
#165 steven mosher:
I wonder if there are enough surveyed sites with acceptable station quality (CRN1, 2, and possibly 3) to test the Oke UHI equation in the USA48.
Here’s my thinking-out-loud procedure:
1. Identity the USA48 stations with the best CRN ratings (there are ~20 CRN1, ~45 CRN2, and ~90 CRN3 stations);
2. Identify a subset of stations that are truly rural (Sr);
3. For all stations in Sr, find non-rural stations within some radius (Su);
4. For all stations in Sn, calculate the annual temperature difference vs neighbouring stations in Su — call these difference Du (ideally Du represents UHI and other urban effects);
5. Shift Du to a suitable reference period that is unlikely to have significant UHI (1890 to 1910 may be suitable for smaller centres), and generate decadal averages — call this Du’
6. Generate decadal population histories for all Su stations using census data (Pu);
7. Plot Pu vs Du’
What do you think?
Step 5 will always fail.
Towns have always had significant UHI. You have to find a period where the population and technology are fairly stable. Even that’s taking a chance.
#174 MarkW:
You’re right — step 5 is a problem.
Maybe the right answer is to take the derivative of Du wrt Pu.
It seems to me we could plot (and analyze) Pu vs dDu/dPu.
Would that work better?
JohnV,
That sounds pretty good.
Population is a pretty good proxy.
The problem is that there are a number of factors that affect UHI. For example, energy useage per person has been rising over the last century.
Another is that the ratio of concrete/asphault per person has changed over the years. Finally, you have to be carefull about comparing towns that are increasing in population vs. ones that are decreasing in population. Most towns are pretty quick to pour new concrete as new residents move int, but they don’t rush out to pull up the concrete when they leave. Concrete and ashphault will break down on their own, but it’s a process that takes decades.
All things considered, population is probably a pretty good proxy for UHI (just remember that it is not a linear function).
It’s possible that all of the things I’ve mentioned don’t make enough difference to matter, but I would like to have that demonstrated with a study, rather than just assumed away.
#176 MarkW:
Those are all important points to consider.
I think it is reasonable to exclude municipalities with decreasing populations. Ideally they would be analyzed separately but I doubt there will be enough stations in this category.
The analysis could also be split into time domains. A coarse split like pre- and post-WWII might be useful.
The purpose of this little study would be to check for and quantify the relationship between UHI population (or log population, or the square root of population, or whatever). To me this is the kind of useful *information* (maybe even knowledge) that can be gleaned from the SurfaceStations *data*. (That is not meant to be derogatory towards SurfaceStations — data is a necessary prerequisite).
http://climatesci.org/2008/02/20/a-new-york-times-report-by-elisabeth-rosenthal-biofuels-deemed-a-greenhouse-threat/
RE 173. I want to get OKE’s paper in my hands. I’ve only read references to it, some interesting expansions of it ( adding other terms to the log(pop). Also, I’ve found some references to how
the Coeff. of the Log10(pop) varies by continent. Conceptually I can accept the Log(pop)
paradigm. ( same with C02). Essentially a thresholding effect.
Since I have no accountablility I would say that UHI is a Function of Building Height
( which is probably a log reponse), Ground cover ( log reponse) and Waste heat.. log(population)
Arm chair analysis, pass the Tv remote.
You wrote
“I wonder if there are enough surveyed sites with acceptable station quality (CRN1, 2, and possibly 3) to test the Oke UHI equation in the USA48.”
I was looking at Orland and Willows last night and have decadal pop figures ( UHI forcing?)
I would like to get my hands on the OKE’s studies, but the best I have been able to do is find
references to his equations. Also, in subsequent work, he also argued that the Coeff of the Log10(pop) vaired by continent. I would hazard this. I would hazard that the UHI response to increased population is likely to be thresholded. Anyway, I have Willows and Orland POPulation figures
for 1880 to present ( decadal) both sites are sub 10K population, but have grown from LT
1000 people to 6000 people over the course of the past 120 years. marysville might be a nice
companion site…
What explains the warming better. CO2 or log(pop)? or combination..hmm
Here’s my thinking-out-loud procedure:
1. Identity the USA48 stations with the best CRN ratings (there are ~20 CRN1, ~45 CRN2, and ~90
CRN3 stations);
I played a bit with this from the other direction.. Without looking at CRN rating
I selected two groups. “best” and “worst” Where “best” was the union of Rural = true.
Nighlghts = 1, GHCN = dark, Brightness index = 0, Open = True. And Worst was the
antithesis of this.
2. Identify a subset of stations that are truly rural (Sr);
Some issues here, but I started to filter sites by the following logic
a. Pop = rural. nighlights =1. GHCN = Dark. Brighness index =0. Site is
open today. I think updating the site with current POP would be a good thing.
Essentially, what we are saying is “identify sites” that have not suffered from UHI.
Actually USHCN has population figures for most of the sites from 1910 on…. hmmm
I also tossed sites on the coast and airports. I Found a couple CRN12 in this batch.
3. For all stations in Sr, find non-rural stations within some radius (Su);
This is backwards of the H2001 approach, but makes no difference. It would be
an interesting excercise, but the fundamental issue is that Urban/Rural is REALY ANALOG
not digital. assume OKE is correct. UHI = .x*log10(P). Assume no global warming.
The TREND in a city that grows from 5K to 15K is greater than the trend in a city
that grows from 100K to 110K people. puzzlement. a growing RURAL would COOL a Urban
site.
4. For all stations in Sn, calculate the annual temperature difference vs neighbouring stations in Su — call these difference Du (ideally Du represents UHI and other urban effects);
I think we want to examine those instances when RURAL cools the URBAN. Since this adjsutment
is unphysical
5. Shift Du to a suitable reference period that is unlikely to have significant UHI (1890 to 1910 may be suitable for smaller centres), and generate decadal averages — call this Du’
Even though UHI was first identified in 1830….
I like this move.. One thing I was playing with was seeing how well Oke’s rule fit
stations that had very shallow population growth…The approach was to find gold standard
sites.. Spatial coverage, of course, suffers with this. However, note how compelling it
is when you select very high quality sites and the answer matches the dogs breakfast,as it were.
6. Generate decadal population histories for all Su stations using census data (Pu);
USHCN has this. its at the FTP. ( lucia should add population as a forcing)
If you need a pointer to the decadal population figures… crap its the USHCN ftp site.
The orginal plan was to use decadal POPS, but ……
Might be cool to update the POP with the UN data. Make sure that Urban 1980 is Urban 2000
7. Plot Pu vs Du’
It would be a start. As Much as I like Oke, other’s have argued that population is
a good start but that other factors explain more. Still, I love data.
What do you think?
I think that GISS is largely correct, but that sites that dont meet quality standards should not be used. Beyond that I think this temperature data is endlessly fascinating.
steven mosher:
A few clarifications and thoughts:
The order of selection (rural first, urban second) is to quickly shorten the list of candidate sites. The truly rural sites (rural, dark, nightights=1, brightness index=0, approved by Anthony Watts, etc) are very few. Starting with them and drawing a fairly tight radius (500km?) quickly shortens the candidate list of urban stations.
I don’t think it’s a good idea to filter by whether RURAL cools URBAN or vice-versa. I think the procedure should be to look at the differences between urban and rural (using stations that we know to be of high quality) and check the strength of the relationship to some function of population.
MarkW pointed out some problems with #5 (shifting) and after thinking about it for the afternoon I’m pretty sure the right procedure is to deal with the derivative of temperature wrt population. Using the Oke equation you gave above:
Du = 0.73*log10(Pu)
dDu/dPu = 0.73/Pu/ln(10) ~ 0.32/Pu
Agreed.
I’m hoping to drum up a volunteer or two to filter the sites and gather the population data. Hopefully Anthony Watts will be willing to review the station lists for problems. I’d be happy to run the calculations.
The recent UHI studies that I’m aware of do not seem overly robust — this could be a real contribution.
#179 I’ve been a fan of Oke’s for years. Will make sure you get the 1973 paper.
re 180, i wasnt even thinging of applying CRN yet, when you sort, rural, nightkight=1, brightness =0
you have 200 sites, very few are crn12. let me see what I can do
Bender and JohnV.. I found this
Click to access O_8_2.pdf
fun stuff
3 Trackbacks
[…] any adjustments to compensate for things like urban heat islands. Places like Cedarville, CA and Tingo Maria, Peru both illustrate some of the oddities with the adjustment methodology used by NASA GISS. One of the […]
[…] “UHI” adjustments outside the U.S. even begin to deal with the problem. Posts were here here here here here […]
[…] Hansen博士声称，不但城市中有温度测量偏高的趋势，而且乡村地区的温度数据竟然还有“偏低的趋势”。Hansen博士于是使用其他城市气象站的温度数据来拉高乡村地区的温度数据！AGW鼓吹者们（以下简称“暖化者”））声称20世纪的温度一共只上升了不到1摄氏度。而根据Hansen博士所使用的怪异调整方法，某些并非乡村地区的气象站，历史温度竟然被他降低了了3摄氏度！ […]