A few readers here at rankexploits have asked questions or made claims about population as it relates to UHI, so I took some time to start a new project looking at population over time at the various stations in the BerkeleyEarth dataset. When it’s done the code will be on github for folks who want to see what I’ve done. Today we will just scratch the surface and depending on your questions ,I may go in the directions that are brought up in conversation.
Population is often considered as a regression variable for UHI. This started with Oke, who related population to the max UHI detected at cities various cities. However, over time, population has been supplemented as explanatory variable for UHI with other variables that relate to the urban energy balance. Population would still play a role here insomuch as it contributes to the anthropogenic heat flux in energy balance approaches; but, on its own, doesn’t do a very effective job of explaining UHI. Other factors matter more.  We can discuss that in the comments; but, for now we will just start by collecting what data there is on population and showing how it varies station to station and how it varies over time.
The source for the data is the Hyde 3.1 dataset, a 5 minute ( roughly 8km ) grid. I’ve compiled the dataset back to 1700. There are various metrics we could use: total population count; urban count; rural count, and population densities. I selected density. The data exists for every decade up to 2000 and then 2005. City size follows Zipf’s law or a power law. And if we look at a histogram of population by grid cell we will find that large cities are very rare and that most of the world is undeveloped. If you want to have some fun reading about this topic  start here for an easy introduction.
I mention this because people are often fooled when commentators select certain cities to demonstrate UHI. They tend to select the large cities, and of course there are very few of those. The vast majority of the planet is unpopulated and as we will see only a tiny slice of station data come from these large cities. The vast majority of stations are not in areas that you would call urban. In the end, however we can just dispense with the terminology “urban†and “rural†and substitute objective metrics that capture the causes of UHI.
To make the data somewhat easier to visualize, I’ve taken the density and re-coded it. To keep it simple I’ve used log of the population density base 10, so on the following map values of 0 represent no population ( 0 to 1 person per sq km) 1 represents 1-10 people, 2 = 11-100, 3 = 101-1000, 4 = 1001-10000, and 5 is 10K+. And we see what we all know: the earth is largely unpopulated.
One concern over UHI is that when we sample the earth we are sampling it generally where people live and thus have a sample that may be infected by UHI. To give you a sense of this I have divided the 43K stations into two classes. Since Hyde data ends in 2005, I defined one set of stations as “Active†if they had temperature data after 2004. Stations that ended before 2004 are “not active†. In the future I’ll develop a population history for every station depending on its start date and end date. But for now, let’s just look at what kind of population density we see at active stations:
A couple of points before we move on. First note that in the active data roughly 15% of the stations are located in areas that have zero population within 5 arc minutes of the station location. Roughly 90% of the data comes from stations that have less than 1000 people per sq km. For reference Hansen defines rural as roughly less than 10 people per sq km.
Those numbers are hard to visualize. To help visualize that I selected random locations in the US that represent zero people, 10 people per sq km, 100, 1000, 10000 and then the max.
Below is a random US location with zero people
10 people per sq km is pictured below
100 people per sq km is depicted below
One thousand per sq km is depicted below. Roughly 90% of stations have population densities less than that.  A couple of points here. If you live in an area where lot size is roughly an acre, and if you have the average number of  people living in your house, then guess what the number of people per sq km that represents? Or if you live in an area with a standard .35 acre plot?
10 thousand people per sq km is shown below. Â One question to ask here is how do we get to density’s this large?
As you can see once we get to the  10K people per sq km we have a landscape that is substantially transformed; and, it’s the transformation of the landscape and not so much the number of people living there that matters. We can discuss that going forward, but for now it’s important to understand that 90% of the sites used by Berkeley Earth are not urban. Whether we use somewhat arbitrary population cutoffs (  1000 people per sq km is suburban) or whether  we use measures of how much the landscape is transformed, the vast major of stations are placed in areas that are not urban. For the most part  you don’t find UHI in the global record because 90% of the stations are not located in areas that have the surface features that cause UHI.
Lastly, below is a picture of the most dense  area in the US
Opps. That was zoomed in on central park. Â Here we are at altitude.
I switched those last two to make a point. When we say urban we have to remember that the urban landscape is heterogeneous, as is UHI. It may matter where exactly the station is located.
Finally, when we talk about UHI studies, the vast majority are done for locations that have densities much greater than 1000 people per sq km. There is very little study of UHI in areas that have population densities less than 1000 people per sq km. And the reason should be obvious by looking at the photo. Those locations are not urban.
BUCL .
BUCL Â Study Area
BUCL  Site  W07
BUCL Â Site W04
BUCL  Site  Wo04  Hires
Paradise Detail
Statistics  for  Non Adjacent Stations.
Stations are scored by their adjacency to urban areas. Here  using BUCL parameters  as a guide. The city center has a population density of  ~5000 people per sq km. Below  the population density is shown for the study area. Paradise, W4 and W7 are shown for illustration.
Filtering stations  only include  A) Active stations in 2005  and B) Stations that are  not  adjacent to  grid cells with populations 5K  or  higher. Adjacency is determined at a 20km  range.
They are not urban, but to they still have a heat island effect?
SM: “There is very little study of UHI in areas that have population densities less than 1000 people per sq km. And the reason should be obvious by looking at the photo. Those locations are not urban.”
.
Assumptions:
1) Unlike urban, rural landscapes are homogeneous — false.
2) Only urban stations can have micro-site issues like a paved walkway or nearby vent grill or HVAC condensor ——false.
3) Cultivation or water management doesn’t simulate UHI — false.
4) UHI causing buildings, drainage and pavement are removed when population has an exodus ———————— false.
5) The technological capacity to produce UHI /per-capita has not changed over the 160-yr transition from pre-industrial —- false.
6) Moving stations previously located on building tops to airports has eliminated most of the bias in the record ————- false.
7) The adjustment for moving a station from urban cools the past but this is compensated for by gradual counter adjustments of the record to account for the gradual UHI past growth—— false.
8) Stations moved to the edge of town no longer experience warming by gradual city expansion ————— false.
9) There is no UHI in the suburbs, only in inner cities — false.
10) There is no good way to study UHI to get a handle on it –false.
11) Studies of UHI are done primarily for concerns of it biasing the GMST record and are often cited in climate science.——— false.
12) Since UHI has been known for many decades it was long ago adjusted for accurately by climate scientists ———— false.
.
Steven, I applaud you for addressing this issue and doing the work for the post as I realize you have been mostly on the other side of these arguments.
Mike
“They are not urban, but to they still have a heat island effect?”
First, we have to understand that the studies of UHI are focused on landscapes that look utterly different than 90% of the station locations.
Let that sink in.
So your question is..
Can we extrapolate what we known about Urban heat islands to areas that are utterly different?
Switch that around.
Suppose I only studied the sub urbs.. and found no sub urban warming effects. Would you think I could suppose that these results could be extrapolated to urban areas? I hazard not.
All that said.. let me answer your question
“”They are not urban, but to they still have a heat island effect?”
First lets establish a boundary
What kind of UHI do we expect to see from a large city.
Oke 1997. for a city of 1 million.. the MEAN value would be
1C to 2C. Note this is for all weather conditions.
Next we can look at the causes of UHI which havent changed in the literature for decades.
I will simplify it somewhat.
There are 4 basic factors
A) changes to geometry
B) changes in surface properties.
C) changes in atmospheric properties.
D) additional Heat from Human activity
Over the years folks have investigated which of these factors matters most– ( lots of papers ) some recnt work on 65 US cities is really interesting, but its SUHI ( surface not air )
A) changes in Geometry. Number one here is Building height.
The higher the buildings the worst the effect ( generally )
Of course population density is a proxy for building height.
You tell me the density and I can predict the average building
height ( somewhat ). Building height is important because
1) the geometry can create radiation canyons
2) the geometry can limit sky view and retard re radiation to space
3) the geometry alters the wind, turbulant mixing and convection efficiency
B) Surface properties. natural surface are replaced with materials that have lower albedo and surfaces that store heat. The key here is the amount of surface that is transformed.
C) changes to the atmosphere. basically aerosols over cities from pollution.
D) human activity. driving cars.. etc
Back to your question
‘Is there a “heat island” effect in these types of areas?
It may matter on whether or not they are close to an urban enviroment. Most of the work you will see is on how the UHI effect drops off as you get farther away from the urban core.. if the city has an urban core.
There may be a “heat island” effect, and if you google heat island images you will find plenty of idealized digrams showing how the UHI decreases as you go away from the city center.
A good question might be how much of the “heat effect” originates in theses low density areas versus how much is transported there.
ANd its important not to forget.. if you are looking for UHI in the global record ( Big effects ) you are not going to find big effects if 90%+
of your data comes from areas with no urban structure..
http://profile.nus.edu.sg/fass/geomr/roth%20uhi%20hefd13.pdf
“Assumptions:
1) Unlike urban, rural landscapes are homogeneous — false.
A) I dont assume that. Quite the OPPOSITE!
2) Only urban stations can have micro-site issues like a paved walkway or nearby vent grill or HVAC condensor ——false.
A) WE ARE NOT TALKING ABOUT MICROSITE YET
3) Cultivation or water management doesn’t simulate UHI — false.
A) Never said that. In fact The OPPOSITE. Water management
is critical. Oke and Grimmond
4) UHI causing buildings, drainage and pavement are removed when population has an exodus ———————— false.
A) Never said that. In fact POPULATION matters Less than
BUILDINGS. See tokyo weekend studies
5) The technological capacity to produce UHI /per-capita has not changed over the 160-yr transition from pre-industrial —- false.
A) Never said that. We are only DISCUSSING THE METADATA
6) Moving stations previously located on building tops to airports has eliminated most of the bias in the record ————- false.
A) Never said that. In actuality the number of stations
on building tops was pretty small and limited to the US
7) The adjustment for moving a station from urban cools the past but this is compensated for by gradual counter adjustments of the record to account for the gradual UHI past growth—— false.
A) Never said that. In fact I have no clue what you are talking about
8) Stations moved to the edge of town no longer experience warming by gradual city expansion ————— false.
A) never said that . We can actually look at that if you like
10) There is no good way to study UHI to get a handle on it –false.
A) never said that.
11) Studies of UHI are done primarily for concerns of it biasing the GMST record and are often cited in climate science.——— false.
A) Never said that
12) Since UHI has been known for many decades it was long ago adjusted for accurately by climate scientists ———— false.
A0 Never said that
Ron. You need to take a breath. Most of the assumptions you try to attribute to me, I either disagree with or would never make.
There is ONLY ONE you got right. and we will address that.
But First, you need to acknowledge that you didnt even expect to see the data you saw today.
1. 90% of the stations are located in areas where the population density is NOT URBAN.
2. Studies of URBAN heat islands cannot be DIRECTLY EXTRAPLOATED to these types of areas.
That is all you need to admit.
1. admit that 90% of the stations are located in areas where the population density is not what you would call urban.
2. admit that the studies you and most people refer to are focused on urban areas..
3. Admit that one cannot SIMPLY extrapolate the results of Urban landscapes to suburban and rural.
Admit those three and we make some progress.
Ron.
The one you got right
9) There is no UHI in the suburbs, only in inner cities — false.
yes I would argue this. There may be a heat island in the suburbs but I would not call it UHI.
My main point is to get people to actually realize that there are only a small number of stations that might suffer from UHI– where by UHI I mean.. URBAN heat island.
We can.. and will.. move on to these lower population places..
However.. when we discuss this 90% we have to be mindful that
results from downtown tokyo, london, NYC.. cannot be mindlessly extrapolated to
A small villages
B) populated rural areas
C) suburbs
So, promise me you wont mindlessly take results of URBAN CORES and apply them to low population areas.
You say 90% comes from places with no urban structure, yet from your pictures I think it is possible there would be some effect though not as much as the big cities of course.
Have studies been done to quantify any heat island effect for class 2 and 3? What percent of the data do they represent?
It seems like you are simply ignoring those classes as ‘not urban’.
Roy Spencer blogged about it
http://www.drroyspencer.com/2010/03/global-urban-heat-island-effect-study-an-update/
Is he way off?
“You say 90% comes from places with no urban structure, yet from your pictures I think it is possible there would be some effect though not as much as the big cities of course.”
1. of course it is POSSIBLE that there is some effect. that is not my point.
2. My point is this.
A) Most if not all of the UHI studies are focused on areas that
have population densities greater than 1000 people per sq km. People just need to ADMIT THIS FIRST.
B) the causes identified in those studies ( tall buildings,
no grass, no trees, etc) are MISSING in the lower population areas.
C) people havent looked for UHI in the suburbs ( there are
some limited studies) because.. the problem is not
as bad there.
D) DONT EXPECT to find big UHI effects… when the vast
majority of sites are not urban.
E) stop trying to understand the problems that may exist
at low population sites, by using data from high population cities.
We dont even have to AGREE on UHI to understand that you are not going to get very far by using studies of cities to understand 90% of a database that isnt taken from urban areas.
“Have studies been done to quantify any heat island effect for class 2 and 3? What percent of the data do they represent?”
Are there studies devoted to cities with sizes of 100 people
and 1000?
generally speaking no. There are some people ( imhoff and
me I suppose ) who have looked at SUHI in the suburbs
and small cities.. but SUHI is different from UHI.. very hard
.. we can talk about that later.
But generally people study UHI in big cities so they can get money to do better urban planning
################################
It seems like you are simply ignoring those classes as ‘not urban’.
Roy Spencer blogged about it
http://www.drroyspencer.com/20…..an-update/
Is he way off?
ya he is way off.
My point right now is not to redo what roy has done, but rather to hammer home some fundamentals. and to give
people a better view of what we are actually working with.
what does 1000 people per sq km look like?
and then..
when you read somebody who posts pictures of temperature stations in a city.. and then talks about UHI in mega cities..
and then.. says the stations are all located in urban areas..
You need to know.
Not really.
First Quiz:
What proportion of all active sites ( 25,000) satisfy the following
criteria
A) The site has less than 1000 people per sq km
B) no grid cell within 20km has more than 1000 people per sq km
and then
What percentage of low population sites ( 1000 ) are within
20km of a high density location ( 10000)?
Pray tell, Mosher, how do you get the log of zero population density?
It has been my view for some time that UHI should be handled for adjustment in the same manner as any micro climate effect that is not representative of the area that a particular station is assumed to cover. Further it must be kept in mind that temperature series from stations and adjusted for micro climate effects are of primary importance in determining trends and the trends are affected by non climate micro climates that change over time. I suspect that micro climate changes of the non UHI type could overwhelm the UHI micro climate effect. While the UHI micro climate effect might have some correlation at some time and in some locations to population, it is not people or numbers of people who change the micro climate but rather what those people build and change and further where the temperature stations are located in that changing (or non changing) environment.
As the Watts effort found in dealing with a well documented snapshot of the micro climate around a station temperature station it tells you very little about the past conditions and when and how fast those conditions changed and thus about effects on the station temperature trends. While meta station data might help in this matter the data is not always reliable or accurate or meaningful for the purpose at hand. All these limitations add up to the fact that adjustment algorithms, like it or not, are the best and least subjective method of looking for these changes and determining the appropriate adjustments.
Several rather different adjustment algorithms are available for comparison and the best known method of making that comparison is benchmarking whereby a reasonable climate free of non climate micro effects is established and then known non climate effects are added and the adjustment algorithms are then applied to a world where the truth is known. In my personal preference I would also like to see slowly changing non climate effects added to test how well these effects (that cannot necessarily be related to any known non climate effect) are detected and adjusted by the algorithms being benchmarked.
Otherwise I suppose we will continue to see a lot of conjecture on these blogs without practical methods spelled out on how to determine the effects of micro climates, like UHI, on temperature trends.
Sky,
LOL!
I guess you assume 0.01 person per square km, or call “less than 1 person per square km” the same as 1 person per square km for calculation purposes.
Pray tell, Mosher, how do you get the log of zero population density?.
######################
Simple.
First you read
“To make the data somewhat easier to visualize, I’ve taken the density and re-coded it. To keep it simple I’ve used log of the population density base 10, so on the following map values of 0 represent no population ( 0 to 1 person per sq km) ”
Then you comprehend
“Oh he is recoding things just for visualization.. and he says
0 represents 0-1 people.
Then you wonder
“how did he handle zero”
Then you look at the chart
“hmm.. all those unpopulated places are coded as zero”
Then you wonder..
“how the hell would he handle .5 people?”
Then you think some more
” Oh ya, he said zero to one was coded as 0″
Here it is easier in code
Locations % mutate(Pop2005 = POP_All) %>%
mutate(Active= ifelse(LateYear >2004,TRUE,FALSE)) %>%
mutate(NeighborPOP=POP_Max) %>%
mutate(LogPop = ifelse(Pop2005%
filter(!is.na(Pop2005)) %>%
mutate(LogPopNeighbor = ifelse(NeighborPOP< 1,0,log10(NeighborPOP)))
############
and then you can also see that we are going to discuss the population of neigbors.
generally speaking when folks take the log of populations we recode anything between 0 and 1 person as 1 person.
But if you to redo the visualization with a more colors and no "coding".. be my guest
“Sky,
LOL!
I guess you assume 0.01 person per square km, or call “less than 1 person per square km†the same as 1 person per square km for calculation purposes.’
For visualization..
“Sky,
LOL!
I guess you assume 0.01 person per square km, or call “less than 1 person per square km†the same as 1 person per square km for calculation purposes.’
For visualization..
Kennth:
############
It has been my view for some time that UHI should be handled for adjustment in the same manner as any micro climate effect that is not representative of the area that a particular station is assumed to cover. Further it must be kept in mind that temperature series from stations and adjusted for micro climate effects are of primary importance in determining trends and the trends are affected by non climate micro climates that change over time.”
1. I would like to keep discussions focused on UHI.
2. If we want to talk about microsite, then we need to define
a physical scale.
3. When people do talk about the micro scale they typically refer to the surrounding 100 meters
4. Understand that the same causes are at play.
“I suspect that micro climate changes of the non UHI type could overwhelm the UHI micro climate effect. ”
1. Its unclear what you mean by non uhi type
2. its unclear what you mean by the UHI micro
“While the UHI micro climate effect might have some correlation at some time and in some locations to population, it is not people or numbers of people who change the micro climate but rather what those people build and change and further where the temperature stations are located in that changing (or non changing) environment.”
1. Yes. The human body as a source of heat is about 100W
not an issue.
2. It is what the people build. Build a parking lot
in the middle of nowehere. put a thermometer close by
and you have microsite. Same cause as UHI.. but not urban
“As the Watts effort found in dealing with a well documented snapshot of the micro climate around a station temperature station it tells you very little about the past conditions and when and how fast those conditions changed and thus about effects on the station temperature trends.”
1. Yes for his study he has to assume that class 1 today
has always been class 1.
I’ll stop with this..because at this point I dont want to get into micro climate or micro site.
In the end.. I think.. that will be the last refuse of the temperature deniers.. Basically.. that
A) features of sites that we dont have data on MUST BE
the cause of warming in the data.
in short the unicorn defense.
But lets not go there yet. First Folks need to understand why UHI is so hard to find… before they run for the microsite defense.
Steve Mosher,
I think you mean “refuge”, not refuse. I prefer ‘the last redoubt’. But yes, some will never accept that adjustments to the historical data sets are accurate, nor even that significant GHG driven warming is real. I am actually a little puzzled that you spend as much time as you do engaging those folks…. they are the same folks who refuse to admit adding 10 gigatons of CO2 to the air each year is what causes the atmospheric concentration to rise. The best advice is what you have often given: do a little science, or at least do some scientific thinking. The problem is (I think) most you tell that to are quite incapable of either. So it is a waste of time to engage them. The endless raging at WUWT and Judith’s over the cause of rising atmospheric CO2, how the temperature record is false, and how back-radiation violates the second law are, well, ….. endless. Better to ignore those who can’t or won’t understand basic physical principles. And that goes for the derranged on the other side of the political divide as well.
SteveF.
ya I get that. For now this is a fun way to finish a project.
and I found some new datasets and literature.. so maybe a paper when I am done.
SteveF,
Typo: it’s 10 GtC per year (roughly), not 10 GtCO2. Otherwise, agreed.
Well, I am having trouble getting past your definition of UHI.
D) DONT EXPECT to find big UHI effects… when the vast
majority of sites are not urban.
E) stop trying to understand the problems that may exist
at low population sites, by using data from high population cities.
We dont even have to AGREE on UHI to understand that you are not going to get very far by using studies of cities to understand 90% of a database that isnt taken from urban areas.
OK, but that doesn’t mean there aren’t problems at low population sites, that is not to be calculated from high population cities.
I will concede that 90% of sites do not come from big cities, but don’t see how that means UHI is therefore not a big issue.
Steven Mosher—Thanks for this. Been out of pocket; now catching up on work. Will look at it more after a student gives his dissertation defense in the morning.
MikeN:
And I don’t see how UHI can be much of an issue given what we know.
I would suggest you propose a testable model where sites in the Artic show the greatest warming (regions with the least anthropogenic activity), and where the effect is largest in the winter, and yet the influence of anthropogenic activity (concentrated at lower Northern hemisphere latitudes) is a “big issue” in terms of its effect on estimated rates of warming.
I would expect if anything that anthropogenic activity in low-population areas to result in a decrease in the rate of apparent warming during summer months and neglible effect on the rate of warming in the wintertime.
“Well, I am having trouble getting past your definition of UHI.
D) DONT EXPECT to find big UHI effects… when the vast
majority of sites are not urban.
E) stop trying to understand the problems that may exist
at low population sites, by using data from high population cities.
We dont even have to AGREE on UHI to understand that you are not going to get very far by using studies of cities to understand 90% of a database that isnt taken from urban areas.
#########################################
OK, but that doesn’t mean there aren’t problems at low population sites, that is not to be calculated from high population cities.
I will concede that 90% of sites do not come from big cities, but don’t see how that means UHI is therefore not a big issue.”
1. UHI is defined as URBAN heat Island, not rural heat island,
not sub urban heat island, not micro site.
2. The causes of UHI have been well known for decades. I will
list them below and explain why it matters.
3. 90% dont even come from cities much less big cities.
4. The perception that UHI should be a big issue stems from
people looking at studies of BIG CITIES or areas with high
population density. Live 7 years in my shoes. 7 years of people
pointing to studies of Seoul, Tokyo, London, Phoenis, etc
and then expressing dismay that we cant find UHI in the
global record. Basically, people come to the study of UHI
( I did ) with this expectation that it has to be big problem.
5. The fact that 90% of the stations are in places that would not
qualify as urban, should cause you to suspend judgement
( ie be skeptical ) that UHI has to be a big problem.
It may be no problem or a small problem, but surely
if Oke determined that a city of 1 Million might see a UHI
of 1-2C.. should raise a question in your mind about the
feasibility of finding a problem in areas that have population
less than 1K per sq km.
For the known causes of UHI, i’ll discuss the two most important
1. geometry. Without tall building packed tightly together
there is no radiation canyon. A radiation canyon forms
when SW incoming is reflected and concentrated. Think
corner reflector. Without tall buildings the Sky view is
wide rather than narrow. With a narrow sky view LWIR
cannot return to space. Without tall building surface
roughness changes and you get different wind flow.
Tall buildings also store more heat during the day.
Building height is one of the key drivers of UHI. consequently
when 90% of the sites are in areas that are not dominated
by tall buildings, we can expect the UHI to be reduced.
2. Surface properties. two factors here Albedo and heat capacity. In short, the UHI you see is related to the sq root
of the impervious surface area. So if you have a sq km
that is all concrete that will produce more UHI than a
sq km that is covered half with grass and trees and
half in concrete.
if an area is lacking the primary causes of UHI, then perhaps people should adjust their expectations of finding a big problem.
That.. is the most important point.. that your expectation of finding a big problem should be dimished. ie you should be more open minded about results that show no problem..
http://cybele.bu.edu/download/manuscripts/peng-uhi-est-2012.pdf
http://www.int-res.com/articles/cr2010/42/c042p209.pdf
https://www.researchgate.net/publication/263543874_Urban_heat_island_and_wind_flow_characteristics_of_a_tropical_city
HaroldW,
You are right, it’s ~10 Gt carbon and 10*44/12 = 36.7 Gt CO2.
Because…
There’s no way to verify that adjustments are accurate.
There’s no way to verify that “significant” GHG driven warming is real.
It’s funny to me that people still make assertions like they know stuff, when its clear they don’t.
Andrew
Heck, climate models don’t even have any scientific requirements.
You could stick your thumb in a pie and call it a climate model as far as I can tell.
Andrew
It’s not people but buidingsvabd asphalt that cause UHI, even a ghost town has an UHI. The population of Vienna decreased since 1900, but the UHI stayed the same.
http://www.tonalsoft.com/enc/v/viennafiles/vienna-population.gif
SM:
.
OK, I agree to put off talking about microsite and land use, land change LULCC but as you point out that since most of the stations are not urban small effects in LULCC have larger proportionate impact on the record.
.
Suburban stations do have UHI, just not as much as inner city, where it is determined to be a health danger. Do you want a citing or do you concede?
.
Most are surprised to learn, as I was, that UHI is almost entirely a nighttime effect, making the health danger alarm seem implausible, (but hey, the alarm is what produces the grant money). UHI carries the daytime heat into the night without actually raising the Tmax significantly. It’s the Tmin that gets raised in the sunset to early morning hours. Come to think of it this poses at time of observation bias by shifting the Tmax statistically closer to midnight.
.
SM:
.
The way I understand the UHI adjustment for a station move is to lower the recorded temperature of the entire interval prior to the move in order to continue forward without a break in the slope of the prior 15-yr trend. If this lowering occurs all the way back to the stations start it is reflecting an assumption that the station at the start had the same UHIE as when it was closed — no added building or population growth. Or is the adjustment phased in after the start in gradual intervals accurately as the UHI likely occurred? You tell me.
“It’s not people but buidingsvabd asphalt that cause UHI, even a ghost town has an UHI. The population of Vienna decreased since 1900, but the UHI stayed the same.”
there are some US examples.. detroit and cleveland.
“OK, I agree to put off talking about microsite and land use, land change LULCC but as you point out that since most of the stations are not urban small effects in LULCC have larger proportionate impact on the record.
#############################
1. HUH? Where do you get the idea that Since most of the
stations are not urban that Small effects have a larger
impact.
.
“Suburban stations do have UHI, just not as much as inner city, where it is determined to be a health danger. Do you want a citing or do you concede?”
I’ve read a few, but I would not concede. So the citation
will be important because I will use the data from that location
So please.. citation..
.
Most are surprised to learn, as I was, that UHI is almost entirely a nighttime effect, making the health danger alarm seem implausible, (but hey, the alarm is what produces the grant money). UHI carries the daytime heat into the night without actually raising the Tmax significantly. It’s the Tmin that gets raised in the sunset to early morning hours. Come to think of it this poses at time of observation bias by shifting the Tmax statistically closer to midnight.
1. Actually it depends on the climate region.
2. In some areas you actually have urban cooling during the day
3. What do the studies say about SUBURBAN “U”HI?
.
The way I understand the UHI adjustment for a station move is to lower the recorded temperature of the entire interval prior to the move in order to continue forward without a break in the slope of the prior 15-yr trend. If this lowering occurs all the way back to the stations start it is reflecting an assumption that the station at the start had the same UHIE as when it was closed — no added building or population growth. Or is the adjustment phased in after the start in gradual intervals accurately as the UHI likely occurred? You tell me.
1. “the UHI adjustment” ? WHOSE UHI ADJUSTMENT???
who the heck are you talking about? which scientist?
roy spencers uhi adkustment? phil jones? hansen?
gallo? karl? peterson?
2. Like I said I have no clue who or what you are talking about
I see better now that Steve Mosher wants to discuss UHI as a separate issue from temperature adjustments. I do not see what value that has but perhaps I should await a punch line.
With the advent of the objective breakpoint approaches to temperature station adjustments I do not see where or why UHI would be handled as a separate issue from micro climate effects and particularly where the interest is in making the station data as representative of the area it is assumed to represent. That non climate effects should be included in the adjustments is a no brainer, but there can also be micro climate effects that are not representative of the area represented that also need consideration.
I hope there is a learning experience taken away from this post and that it is not merely being used to personally put down individual skeptics and then make generalizations to skeptics as if they are a homogeneous group.
Kenneth Fritsch:
I’m not entirely certain they should be called “objective breakpoint approaches.” There is always a degree of subjectivity to them. For instance, there are at least half a dozen parameters one could tweak in the BEST’s empirical breakpoint calculations. BEST has never explained why they chose the values for those parameters that they chose, nor has it explained what effect choosing different values would have, but it is fairly easy to see the choices are not “objective” and they does have significant effects on the results.
I do think handling UHI via breakpoint analysis may be a good idea (depending on if one’s approach to finding them can detect UHI signals), but we should keep in mind breakpoint calculations are effectively fitting models to the data. It isn’t objective.
And depending on the model being used, some forms of non-climatic influences may be impossible to detect.
OK, so you have limited UHI to big cities, and then declared that it is not a problem because this is just a small portion of the overall record. Note my very first question didn’t say urban heat island.
1) Could 1-2C in 10% of the record give you .1-.2C in global record?
2) What share of stations are class 2 and 3?
3) If you have a lesser heat island(say .5C) effect in class 2 and 3 that are covering more stations, that is still substantial. I don’t see how it is helpful to declare UHI is urban only, forget the rest, unless you are saying the amount of heat island is very small.
“I’m not entirely certain they should be called “objective breakpoint approaches.†There is always a degree of subjectivity to them. For instance, there are at least half a dozen parameters one could tweak in the BEST’s empirical breakpoint calculations. BEST has never explained why they chose the values for those parameters that they chose, nor has it explained what effect choosing different values would have, but it is fairly easy to see the choices are not “objective†and they does have significant effects on the results.”
The parameters, as I’ve explained, were tested systematically. That is, varied across a wide range of reasonable values. That actual change point detection code has two parameters: max segment length and a probability threshold.
If you download the code and install it, there are a series of options for the breakpoint code.
Folks need to distinguish between TWO different processes:
Slicing
Weighting.
Slicing is essentially adjusting metadata. A station has an identifier XXX. If the station moves.. it is treated for what it is.
A new station. Change in TOB.. new station.. Gap in data… ( thats a choosable variable) new station.
Then comes empirical break points. you can turn it on or off and see what happens. report your results and you are welcome, glad the code could answer questions you had.
Slicing and breaking has zero effect. you are telling the program that this segment or station can be treated as a separate entity.
so you start with about 40K stations and after slicing you have 180K or so. ( I’ll double check.. working from memory )
The next thing is weighting. and here too you have some tunable
parameters.
Testing consisted of doing sensitivity analysis of the choosable parameters and then seeing what effect if any they had on the global result. The only parameters that have any impact are
basically turning adjusting on and off and turning emprical breakpoints on an off.
For example. If we use Nic Lewis’ base and final periods
as a test.. we see that fully adjusted is +.16C over no adjustments. and empircal breakpoints are about half of that.
So raw gets you X
No emprical cuts gets you X + ~.08
All adjustments gets you X + .16
So you can fiddle with the options in empirical.. and move that
.08 around.
After publishing we made one major change to the approach after we tested against the benchmark database and improved our skill.
There is some evidence that we dont go far enough with adjustments. That’s some current research, but if you compare the “blinded” approach or what Kenneth calls objective.. with a human assisted approach, the series we adjust “objectively” tend to be cooler than the human assisted expert approach.
Bottom line: when it comes to questions that have policy relevance.. the difference between raw and adjusted.. using Nic Lewis’ approach.. indicate that adjusting land moves the TCR by about 5%.. given the larger uncertainty in that calculation.. fiddling around with adjustments remains a technical curiousity.. fun but no headline results. Put another way, its useless but fun to focus on. And the evidence from comparisons to other approaches suggests we might be low. That somewhat misses the real point of why we did it the way we did.
The issue that skeptics had raised was that human approaches were biased high.So, we wanted an approach that was as data driven as possible. And yes, human make decisions here too, but the decisions are largely insensitive.
The other way to see this is to just use the unadjusted stations.
There are 15K stations that are quality rated highest and dont get adjusted.
Lastly you wont find in the publications the results of the testing.. The code is there for folks to install and run and
improve. pick a parameter. change it. run the code. test it against
the benchmark database. If you improve the skill, write me an email. Last month, one guy submitted some code for improving the regression.. after review with outside folks ( other people who do temperature series.. I asked them what they thought) the submitter withdrew his submission because it ended up being wrong.
The reason for posting code is so that others can ask the questions that
A) the team found uninteresting
B) the reviewers found unintersting
C) nobody thought of
D) received incomplete attention in the publications.
There will always be incompleteness in a record. But if you post the tools for people to answer their own concerns, those with legitimate science concerns ( like the researcher who submitted code), then you have a pathway to better science.
here Kenneth
function [breaks, prob_list] = changePointDetection( data, threshold, segment_max_length )
if length(data) < 6 breaks = [];
prob_list = [];
return;
end breaks = [];
if nargin < 3 segment_max_length = 12*10;
end if nargin < 2 threshold = 0.999;
end data = data – sum(data)/length(data);
threshold_mult = erfinv( threshold ) * sqrt(2);
probability = ones( length(data), 1 );
p_last = 0;
s_limit = 0.001^2;
done = false;
while ~done
s = sort(data);
lower_limit = s(ceil(length(s)*0.1));
upper_limit = s(floor(length(s)*0.9));
mask = ( data lower_limit );
data2 = data(mask);
indices = 1:length(data);
cs = cumsum(data2);
cs2 = cumsum(data2.^2);
len = length(data2);
TO = zeros( length(data), 1 );
T = zeros( len, 1 );
for k = 6:len – 6 m1 = cs(k) / k;
% mean( data1(1:k) );
m2 = (cs(end) – cs(k)) / ( len – k );
% mean( data2(k+1:end) );
% s1 = mean( ( data2(1:k) – m1 ) .^ 2 );
s1 = ( cs2(k) – 2*m1*cs(k) + k*m1^2 ) / k;
% s2 = mean( ( data2(k+1:end) – m2 ) .^2 );
s2 = ( (cs2(end) – cs2(k)) – 2*m2*(cs(end)-cs(k)) + (len-k)*m2^2 ) / (len-k);
% Avoid edge affects by assumming a common population if k < 20 && s1 < s2 s1 = s2;
end if len – k < 20 && s2 p_test );
probability( fx ) = p_test( fx );
[~,fk] = max( TO );
if TO(fk) > threshold_mult k = fk; breaks(end+1) = k;
p_last = probability( k );
f2 = find( breaks > k );
if isempty( f2 ) f2 = length(data);
else f2 = min(breaks(f2));
end
f1 = find( breaks segment_max_length len_seg = breaks(k+1) – breaks(k) – 1;
midpt = floor( breaks(k) + len_seg / 2 );
select1 = breaks(k):midpt;
[breaks2{k}, prob_list2{k}] = changePointDetection( data( select1 ), threshold, … segment_max_length );
breaks2{k} = breaks2{k} + breaks(k) – 1;
prob_list2{k} = (1-prob_list(k))*(1-prob_list(k+1))*prob_list2{k} + … (prob_list(k) + prob_list(k+1) – prob_list(k)*prob_list(k+1));
select2 = midpt+1:breaks(k+1)-1;
[breaks3{k}, prob_list3{k}] = changePointDetection( data( select2 ), threshold, … segment_max_length );
breaks3{k} = breaks3{k} + midpt;
prob_list3{k} = (1-prob_list(k))*(1-prob_list(k+1))*prob_list3{k} + … (prob_list(k) + prob_list(k+1) – prob_list(k)*prob_list(k+1));
end
end
PP = [prob_list prob_list2{:} prob_list3{:}];
[breaks, I] = sort( [breaks breaks2{:} breaks3{:}] );
prob_list = PP(I); breaks = breaks( 2:end-1 );
prob_list = prob_list( 2:end-1 );
Steven Mosher writes:
Don’t worry folks, we’ve tested it. We haven’t published the results of the test or even discussed the tests in any meaningful way, but… we’ve tested them. Trust us! We’re the BEST! We’re completely open and transparent!
Mockery aside, I should point out this doesn’t address the point I made, that these methodologies are not completely objective. Even if one does tests to try to determine what parameter values should be used, there will still always be subjectivity to the matter. This is a simple point, and I’m not sure why Mosher wouldn’t acknowledge it. Then again, he also writes:
Which is either false or worded in a way so as to be intentionally misleading. One obviously could not do the BEST changepoint analysis with just the two parameters Mosher lists here. Similarly, Mosher writes:
But anyone who understands how using breakpoint analysis to detect discontinuities knows slicing/breaking longer series into shorter ones does in fact have an effect. That is the entire point of doing it.
I don’t know if Mosher is trying to rely on some sort of semantic trickery where he can defend statements like these as “not wrong” because of some subtle nuance of word choice or details he intentionally excludes or if he just has no idea what he’s writing. Either way, it’s a waste of time.
“I see better now that Steve Mosher wants to discuss UHI as a separate issue from temperature adjustments. I do not see what value that has but perhaps I should await a punch line.”
ah.. the road map.
Lets classify some of the issues with global temperatures.
1. anti scientific objections to the whole enterprise.. ie
global temps dont exist.
2. Questions/doubts about extrapolation.
3. Questions about bias
A) UHI Bias
B) Microsite bias
C) adjustment bias.
Just practically here is what happens.. If you start talking about adjustments.. people will.. go in various directions and invariably at some point say “What about UHI” that is when you explain the adjustments.. or point to the tests.. or give them the code
they will at some point say what about UHI?
And if you are talking about micro site.. they will launch off into adjustments or UHI..
It used to be that there was a great zombie argument about the great thermometer march.. Chiefo and McKittrick started that one and it took years of focus to kill that zombie..
So.. in my mind there are two zombies left..
The UHI zombie and adjustment zombie..
Micro site? a real concern I think..
So I would rather dispatch with the UHI zombie first. that may take a few years.. haha.
Along the way we may gain some insights into micro site and adjustments.. but for now I’d be happy if the final result is this
People come to recognize that the record is not dominated by
what the literature calls UHI.
one zombie at a time
“Which is either false or worded in a way so as to be intentionally misleading. One obviously could not do the BEST changepoint analysis with just the two parameters Mosher lists here. Similarly, Mosher writes:”
The changepoint detection code has two paramters. The code is posted above.
I post the code to AVOID the semantic games that Brandon plays.
Upstream there are dozens of parameters. For example.
You have to look for breakpoints between neighbors. Well how many paramters determine neighbors…more than two
But then upstream of that are the choices to include or exclude stations… More parameters.
The point is to avoid semantic games ( brandon calls is break point analysis.. ) you have to be more precise.
So detection? there is code.. it’s posted above.. you can work from that back upstream
“Mockery aside, I should point out this doesn’t address the point I made, that these methodologies are not completely objective. Even if one does tests to try to determine what parameter values should be used, there will still always be subjectivity to the matter. ”
Its not simply a matter of subjectivity versus objectivity as those two terms are not very precise. Did a human make a choice?
yes. for example.. you choose the threshold at .95. If folks want to call that subjective I have no issue. A human picked it.
But if you test .8 to .99 and find no substantial difference what do you call that? well you could say.. picking the range is subjective.
The problem with the word “subjective” is that for many people they will assume that the selection was motiviated by some nerfarious intent.
“Don’t worry folks, we’ve tested it. We haven’t published the results of the test or even discussed the tests in any meaningful way, but… we’ve tested them. Trust us! We’re the BEST! We’re completely open and transparent!”
There is a ton of stuff that was tested that isnt a part of reports.
A ton of UHI work, work on the ocean, work on the regression and work on alternative assumptions for adjustments. The most effective way for people to understand the impacts of all this is to run the code. that is why we provide it. we provide it because you cant write up every finding, every test and actually get it published. Bottom line.. the difference between raw and adjusted is on the order of 15%. In between those two values there are an infinite number of results you can obtain by making different assumptions. So.. is the work open? yup. transparent?
yup? exhaustive? nope. you can fiddle.. got get the code and fiddle.. but chances are you wont find anything that merits a scientific finding that has some sort of relevance to the debate.
“But anyone who understands how using breakpoint analysis to detect discontinuities knows slicing/breaking longer series into shorter ones does in fact have an effect. That is the entire point of doing it.”
No breaking series doesnt have any effect.
there are two steps.
the first step is “slicing” now slicing isnt a very good description because it makes you think that something is done to that data
You have a series
A) 0 0 0 0 0 0 0 1 1 1 1 1 1
Your breakpoint analysis tells you that this might be two stations
A) 0 0 0 0 0 0 0
A1) 1 1 1 1 1 1
That is what slicing does.
Things to note… the data doesnt change. You can take that data and run it through nick stokes algorithm for example. All you are doing in slicing is saying “THIS is a different station”
The next stage is weighting.
So you can slice without weighting.
if you slice and dont weight… then you will have ZERO effect.
because slicing does nothing.
What slicing does is this
I can now compare both segments independently.
A0 will be compared to the field. Perhaps it doesnt differ from the field. its weight will be one.
A1) will be compared to the field. depending on how and if it differs it will get a weight.
Depending on all the neigbors weights the weight of A1 will be re adjusted and re adjusted until the error is minimized. So its weighting not slicing.
So. slicing in and of itself… doesnt change data values..
Slicing makes it possible for the weighting function to individually weight a segment.
All that said.. I’ll do a special post for brandon ( hey sorry you lost your work ) on adjustments.. But for now we should focus on UHI.
cause that is the topic.
Wow. Six comments in a row, five of which respond to my single comment. That seems unnecessary. I tried to read them all, but I gave up. There are too many things like:
This repeated claim of me playing semantic tricks follows from me having written:
When I referred to there being many parameters tied to BEST’s “empirical breakpoint calculations,” Mosher decided to say there are only two parameters used in the final step of the calculations, “segment length and a probability threshold.” Neither of these parameters account for things like how one determines which stations to compare for the breakpoint analysis. You need parameters to account for things like how far away stations being compared to try to find breakpoints can be from one another and if you wish to weight them differently based on that distance.
I referred to the calculations used in determining breakpoints as a whole because they all influence where and when breakpoints are assigned. Mosher claims this is me playing semantic games because if you only look at the final step of the calculations, only two parameters are used. Of course, the last step of one’s calculations is not the entirety of one’s calculations, Mosher’s interpretation of my words is clearly not what I meant.
As I said, it’s a waste of time. Everything I wrote is correct, but I don’t see any value in dealing with Mosher’s incessant insults and obvious misrepresentations to demonstrate it.
Brandon
“When I referred to there being many parameters tied to BEST’s “empirical breakpoint calculations,†Mosher decided to say there are only two parameters used in the final step of the calculations, “segment length and a probability threshold.†Neither of these parameters account for things like how one determines which stations to compare for the breakpoint analysis. ”
I specifically referred to this as breakpoint DETECTION
You talk about “empirical break point calculations”
well that process is super long.. in terms of parameters there are more than the half dozen you claim
so for KENNETHS benefit.. I start at the end.. with the ACTUAL CODE..
Not your sloppy way of talking about things
but the actual code.
The detection is two parameters.. three if you look at the hard coded number for shortest segment.
Next UPSTREAM comes the selection of data.. That involves more parameters..
Next UPSTREAM from that are even more parameters
Now, if you want to be specific, you have the code. please refer to it properly and like last time you had questions about code please do a little research before asking me about R to matlab stuff.
Now, if you can wait for your own special topic rather than hijacking yet another topic that would be cool.
I’m done. I’m done. I just don’t even… anymore. Steven Mosher accuses me of playing semantic games because Kenneth Fristch wrote:
And I responded:
But as he explains:
It’s too funny. Who jumps into a discussion then gets snippy over people not talking about what he’s talking about? And who says:
Because a person said there “there are at least half a dozen parameters”? It’s too much. Apparently my “sloppy way of talking things” by saying there are at least half a dozen parameters was super sloppy because there are really more than six parameters?
“There are at least six parameters that could be tweaked.”
“YOU’RE WRONG! THERE ARE MORE THAN SIX PARAMETERS THAT COULD BE TWEAKED!”
I hope everyone will forgive me for “hijacking yet another topic” with my semantic games and sloppy talking.
Andrew_KY,
“It’s funny to me that people still make assertions like they know stuff, when its clear they don’t.”
.
You on the other hand I doubt you will ever need to make any assertion about knowing anything. That you seem to think this is a good thing is very, very strange.
.
You can choose to contribute to a technical discussion, or you can choose to make yourself irrelevant. So far, it looks to me like the latter. When someone as tolerant as Lucia thinks your comments need to be moderated, you should take heed. I doubt you will.
SM:
.
1) If you can forgo the formal credit of publishing you can post any documentation that you create now-a-days. I realize this could give fodder to those that would like to make you look foolish by finding your team’s mistakes but this is the trade-off of claiming transparency. Just because politicians created a fashion of declaring their themselves transparent while breaking the law in their degree of secrecy to avoid vulnerability to “the enemy,” does not mean it’s a good fashion to follow. That said, you are exceptional in your degree of engagement here and I very much respect that on its own.
.
2) The 15% difference between the adjusted data trend and the raw trend is not the universe of possible bias. The true potential bias includes not only wrong adjustments but the neglect of those that should have been made or their incorrect implementation, LULCC and non-inner-city UHI, for example.
.
.
Here is a 100-plus page dissertation from CUNY in 2014 studying SUHI and lot size in New Jersey. Their concern is the danger to the environment that SUHI poses and point out the “climate change” is expected to make it more dangerous. I can’t fault their motivation or their future grant funding. It just seems ironic that we are using their data (or not) to demonstrate the potential oversights in “climate change” claims.
http://academicworks.cuny.edu/gc_etds/28/
The full paper is in the HTML reader on its page
.
Also from Parker 2010:
.
It just makes sense, right? Do you concede on this?
Ron. I already read the dissertation.
Please note that it concerns suhi.
So it’s not very helpful because suhi as measured by midis is just a transform of land class as detected by another midis product.
Wrt Parker is that the example you want to use to define suburban?
You realize that suburban in the uk has a much highet density? Right?
Wrt the dissertation one good thing in it is reference to the Rpa. Agreed?
Steven can we agree that SUHI is to UHI what pre-hypertension is to hypertension or what a second degree burn is to third degree burn?
.
I just got back from a walk on this warm night and could clearly notice the cooling in ambient temp from stepping off the road to the lawn and warming again near the house front.
.
Do you agree that the universe of possible bias exceeds the delta of adjusted minus raw trend?
.
Do you agree that BEST’s slice and weigh method could have bias that could under-weigh non-GHG local warming?
.
Do you agree that UHI, with its delay of Tmin, introduces a positive TOB for midnight demarcations and thus changes the TOB corrections to midnight as well? Is this acknowledged or accounted for by Hadley, GISS or anyone?
.
Do you have an links to databases listing all the Long and Lat coordinates for GHCN stations? Past and present would be ideal.
.
Rpa? Is that remote sensing population assessment?
How does GISTEMP do their UHI adjustment?
Has anybunny ever done a paired site study where one or more urban site anomalies were compared with rural and/or per-urban sites some distance away but which were close enough to “share” the same weather patterns?? Sort of CRN before CRN
Steven, do you agree with the dissertation and Parker 2010 that UHI and AGW pose an urban health threat alarm? Parker 2010 final sentence:
.
Do you feel that the public is being transparently informed that UHI mainly warms winter nights and that AGW also warms winter nights over summer days and favors uninhabitable tundra and sea ice over tropical or temperate latitudes? If so, is this accidental? Can intentional misinformation to a domestic populace in peacetime be justified for a greater good?
Eli, here is Hamdi(2011) and Kim and Baik(2002)
Have you accounted for the suburban special heat island of Philadelphia, California, and particularly Alaska?
“Steven can we agree that SUHI is to UHI what pre-hypertension is to hypertension or what a second degree burn is to third degree burn?
No. There are three things youu need to realize about SUHI.
1. ) the air temperature above the land ( say at 2m) is not
very tightly coupled to the surface temperature.
2. ) they actually dont measure SUHI with Modis. I spent
months doing LST work> I’ll suggest you read the ATBD
for LST ( I’ll link it for you if I can find it)
3. the accuracy is pretty bad.
.
I just got back from a walk on this warm night and could clearly notice the cooling in ambient temp from stepping off the road to the lawn and warming again near the house front.
A) yes there is microsite. we are not discussing that
B) you would be surprised how little the air 2m above asphalt
is warmed. If you search through urban planning docs
and studies from paving companies you’ll be surprised.
There are a couple of published studies that are easier
to find.
.
Do you agree that the universe of possible bias exceeds the delta of adjusted minus raw trend?
I will have to think about that, but I dont think so since
Raw rural isnt that much different than other series..
.
Do you agree that BEST’s slice and weigh method could have bias that could under-weigh non-GHG local warming?
Looking at unadjusted data that has not been sliced or
weighted… I’d say the bias is +- 2%
.
Do you agree that UHI, with its delay of Tmin, introduces a positive TOB for midnight demarcations and thus changes the TOB corrections to midnight as well? Is this acknowledged or accounted for by Hadley, GISS or anyone?
1. We are not discussing adjustments, but no.
2. Hadley and CRU accept ingested adjusted data as is.
.
Do you have an links to databases listing all the Long and Lat coordinates for GHCN stations? Past and present would be ideal.
The data is at the noaa ftp.
.
Rpa? Is that remote sensing population assessment?
1. Did you read the dissertation? Maybe not. If you had I would
have expected you to jump up and down about the various
LST for the zones established by the RPA
Ron.
Here is the ATBD
http://modis.gsfc.nasa.gov/data/atbd/atbd_mod11.pdf
“How does GISTEMP do their UHI adjustment?”
it’s pretty simple. covered at climate audit.
however, in the end it doesnt make much of a difference.
This post is not about adjustments.
Also Ron, we can discuss SUHI separately.. you’ll first want to google
Estimating air temperature from LST
Once upon a time i was assisting some guys who wanted to do a better job in arctic using LST data.
monster sized data.. very knarly problems
“Has anybunny ever done a paired site study where one or more urban site anomalies were compared with rural and/or per-urban sites some distance away but which were close enough to “share†the same weather patterns?? Sort of CRN before CRN”
Some of the work Zeke and I did probably qualifies ..
If we can keep these guys focused on a definition of urban I can probably do something for them to show them what the raw data shows..
“Steven, do you agree with the dissertation and Parker 2010 that UHI and AGW pose an urban health threat alarm?
We are discussing population and urbanity.
You owe me an answer about whether you wanted to accept London suburbs as a definition
“I hope there is a learning experience taken away from this post and that it is not merely being used to personally put down individual skeptics and then make generalizations to skeptics as if they are a homogeneous group.”
The simple way to avoid this is.
A. If you see yourself as a skeptic, then call out bad arguments
when other skeptics make them.
But if you identify as a skeptic and then remain silent when other skeptics say things you disagree with or know to be wrong, then dont be surprised if you are grouped with them.. because that is what your silence does.
So. for example, when I see Cook publish junk.. I say.. that’s junk.
here Ron
these will help you make a better argument
http://www.mdpi.com:8080/2073-4433/1/1/51/pdf
http://hummedia.manchester.ac.uk/schools/seed/Architecture/research/csud/city-weathers/programme/Grimmond_ClimateOfLondon.pdf
Not sure if you want o use Jones and Lister.. remind folks here what they found about UHI in London
you probably want to read the follow up
http://onlinelibrary.wiley.com/doi/10.1002/wea.679/pdf
Paradoxically, London’s nocturnal heat
island was actually strengthening during
the decades of declining population.
During that time, though, there was a substantial
increase in the number of cars.
However, the counter-trend is further evidence
of a temporary synoptic signal
superimposed on a heat island in central
London that was largely established by
earlier phases of urban development.
Jones and Lister (2009) report that urbanrelated
warming has shifted from central
locations to the periphery of London. This
is consistent with observed behaviour of
the UHI in other global cities. For instance,
long-term temperature data for Toronto
show that the most recent increases in
temperature are at urbanizing suburban
stations (Mohsin and Gough, 2009). Recent
warming in Japanese cities is largely attributed
to regional temperature rises, but
the urban signature amplifies where there
is greatest population density (Fujibe,
2009). Clearly, any long-term changes in
UHI behaviour should be interpreted on a
case-by-case basis.
and
http://metlink.org/wp-content/uploads/2013/07/urban_heat_island_-_manchester.pdf
McMurdo Station is the largest base in Antarctica, with some 1000 people in the summer and just under 200 during the winter. A few kilometers away is Scott Base, run by New Zealand. The number of personnel there is more than an order of magnitude lower, 85 in the summer and under 14 in the winter.
GISS supplies temperature data for both locations, and the data overlap from 1957-2008. Using the annual temp numbers I calculated the slope of the data for both stations:
McMurdo 0.24°C/decade
Scott 0.15°C/decade
The warming trend at the larger station, a couple of kilometers away, comes in at 60% greater than at Scott.
DB,
How much have the populations of the two bases increased over that period?
Steven Mosher (Comment #149356)
I think there is a great waste of time “calling out” individuals on these blogs where the discussion can devolve into personalities. Here I include calling out people on any of the sides of the AGW discussion. It is much better to follow a discussion that analyzes and critiques a paper, an article or a post without getting into personalities.
In the discussion above between you and Brandon, you both would do better to present your ideas and views without going out of your ways to discredit others. I tend to agree with the essence of a number of both of your observations and it is much easier to get that essence when they are made without involving personalities and the arguments getting lost in those exchanges.
I should have said about breakpoints that that approach is less subjective than using exclusively meta data. These adjustment algorithms will never be devoid of some subjective choices and that is why we need sensitivity testing of those choices and/or a benchmarking test as I noted above.
Steven Mosher (Comment #149357)
Ron and Steve, I think those authors in the first linked post might do better in their reasoning to consider that the land to ocean trend divergence is related to the land warming (cooling) rate.
Thanks to Steve for putting up this essay on population and UHI
Previous efforts were Anthromes and UHI in 2012 and Zeke in 9/392010 at the Blackboard.
–
To avoid confusion Steve meant to point out there are are no 43K stations in the BEST GHCN figures that he used above.
There are less than 2500 active stations currently 2013.
There have only ever been 7280 initial stations spread over 1700 to now.
Thanks to splicing the number of individual stations, average life 5.9 years , <4% over 30 years grew to about 47,000.
–
The assertion that 90% of stations are rural and 10% urban could be challenged (more sites in cities) but the basic effect is that even if urban sites had 2 c increases it would make no real difference to average global temperatures as Steve said
–
The one confounding point he ignores however is the admitted (by BEST) biases in the temp record.
Their global average land temp was 8.9 C with an error range of 1.6 C, incredible.
–
The problems lie in their distribution and weighting of temperatures. Longer records with more dense spacing and reliable instruments i.e. Urban, are given much greater weighting than sparse irregularly reporting poor instruments Rural.
Further the jacknifing technique blends these records into the nearest 8 surrounding records and modifies them upwards in a spreading ring of Algorithm,not instrumental UHI .
The technique of modifying outliers by weighting away 96% of their value removes all cold rural records.
Zeke admits UHI exists.
BEST and Steve cannot find it when they look at all the records.
How can they?
UHI is not a city problem it is a BEST procedural problem and invisible due to the very means they use for temperature adjustment.
.
The military, which sends human beings to very hot places from time to time, has perfessers looking at heat stress.
.
When calculating heat stress, an important factor is period of acclimatization. That’s typically about two weeks, and that’s about what the football folks have concluded as well as the CDC.
.
So, ironically, the longer intense heat events last, the less adverse they become, at least from direct heat stress.
.
Both AGW and UHI have been and are, in comparison, very slow processes of periods much longer than two weeks, so it wouldn’t appear that either is a significant direct health risk. ( of course, the average American, and presumably developed world citizen, spends 7% and falling of their time outdoors anyway ).
.
Now, acute heat waves can be more brief than two weeks, but those events are by definition weather, not climate. And weather has always occurred and will continue.
.
And, paradoxically, there are some reasons to believe that heat waves would decline with global warming, or at least not increase in step with average temperature increase.
.
First, part of AGW is water vapor feedback. Increasing the average latent heat content of the atmosphere may have the effect of reducing temperature variance. That’s what Manabe and Wetherald concluded from very early GCM experiments. “the variance of temperature in the lower model troposphere decreases substantially in response to the CO2 increase.”
.
Second, acute heat waves occur because of stagnant mid-latitude summer anticyclones. Again paradoxically, these anticyclones are cold polar air masses which enable subsidence above them. When these masses reach the mid-latitudes and stagnate during summer, the shallow layer beneath the subsidence accumulates sensible heat from the subsequent sunny skies. There are many factors, but it’s not clear that an AGW world would produce the same intense polar air masses which are the necessary precursors to heat waves, in view of Arctic Amplification.
.
Third, the subsidence from the events above means the lowest shallow layer of the atmosphere is decoupled from the atmosphere above via turbulent exchange. The heat wave doesn’t have access to the sensible heat of the rest of the troposphere. Radiative exchange does still take place, of course. But the relevant radiative exchange is no longer that at the tropopause, but of the lowest layer. The RF for the lowest layer during heat wave conditions from a CO2 doubling tends to be less ( though not zero ) than the RF at the tropopause.
I have a dumb question.
I have been reading about UHI for years and still don’t understand what the issue is and why it must be adjusted for.
Is it because it is warmer downtown versus the rural area around the city? I don’t think so, because the anomaly data should take care of that – right?
If it because of the baseline – that downtown Minneapolis might have a greater population now than in 1951 or 1961 or whatever, and therefore it is warmer today than 1951 or 1961 and we want to adjust for that?
I guess my question is what is the adjustment and why do it?
If it is warmer downtown Chicago today than at some point in the past or even warmer in downtown Chicago today versus rural today – isn’t that the temperature? Isn’t the instrument reading correctly?
I apologize if this is a stupid question – but I really need a basic answer to explain what the big deal is about UHI.
Put another way – if the correction is to adjust for changing population in an area over time – why do we need to adjust for that?
Ron Graf:
It’s worth pointing out that ~15% difference in trend between adjusted and unadjusted BEST data only applies to its global results. If you look at results for smaller areas, even just at continental scales, the numbers can get higher. There are even areas which have significant trends added to them.
Another factor to consider is spatial resolution. If you average/smooth your data out enough, you can reduce all the globe to a single, global record. That’s useful for some thing. It isn’t useful for other things. Sometimes you want to look at things on a finer scale. Adjustments to the data can affect one’s ability to do that. In the case of the BEST data set, their adjustments greatly reduce the resolution of their results.
That is of particular interest when looking at things like UHI. Obviously, if you smooth your data out enough, it will no longer be meaningful to compare “urban” and “rural” locations because any possible signals would have been too smeared together to separate out. That means “adjusted” results which show urban and rural locations have the same trends don’t actually demonstrate any effects from things like UHI have been removed.
(There are, of course, other approaches which could be used to make such a determination.)
“I should have said about breakpoints that that approach is less subjective than using exclusively meta data. These adjustment algorithms will never be devoid of some subjective choices and that is why we need sensitivity testing of those choices and/or a benchmarking test as I noted above.”
I would be wary of terminology like objective and subjective. It’s better to just describe what was done. And yes, you cant avoid making choices. For example we choose to calculate break points by looking at temperature data. That’s a choice. Is it subjective to choose to look at temperature data.. of course I knew what you meant by objective.. but its easy to twist that and say… oh they made choices.. its subjective.
“To avoid confusion Steve meant to point out there are are no 43K stations in the BEST GHCN figures that he used above.
There are less than 2500 active stations currently 2013.
There have only ever been 7280 initial stations spread over 1700 to now.
Thanks to splicing the number of individual stations, average life 5.9 years , <4% over 30 years grew to about 47,000.
1. No wrong.
2. GHCN MONTHLY DATA has 7K stations ( and growing )
3. GHCN DAILY has 75,000 stations
https://www.ncdc.noaa.gov/oa/climate/ghcn-daily/
but not all of those stations have temperature data..
ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/daily/ghcnd-stations.txt
4. here are some of the other sources
https://www.ncdc.noaa.gov/data-access/land-based-station-data
The other sources FSOD and GSOD also 10s of thousands of stations
This is just basic
“The one confounding point he ignores however is the admitted (by BEST) biases in the temp record.
Their global average land temp was 8.9 C with an error range of 1.6 C, incredible.”
No wrong again.
That is the estimated 2 sigma error in a SINGLE STATIONS MONTHLY estimated temperature.
“The problems lie in their distribution and weighting of temperatures. Longer records with more dense spacing and reliable instruments i.e. Urban, are given much greater weighting than sparse irregularly reporting poor instruments Rural.”
Ah no. wrong again.
1. There are two kinds of weighting. Quality weighting and spatial
weighting.
2. Quality weighting.. the approach knows nothing about the
length of a series or whether it is urban or not.
Quality weighting weights segments of records
3. Spatial weighting. Dense stations are not given more weight BY DESIGN.. that it is what all spatial interpolation does. In simple versions its just IDW.
“Further the jacknifing technique blends these records into the nearest 8 surrounding records and modifies them upwards in a spreading ring of Algorithm,not instrumental UHI .”
Wrong again.
jacknifing does not blend records. The jacknifing is used to estimate uncertainties. I think you read a JeffId post and didnt understand what he was saying.
That said, we tried Jeff’s suggestion and the error bars DECREASED.
“The technique of modifying outliers by weighting away 96% of their value removes all cold rural records.”
wrong again.
if you leave outliers in the record is actually WARMER
But its funny you think there are cold rural records
Let me ask you
What criteria did you use to decide something was rural
That is the topic of this post. Now, it seems you have been hiding a rural defintion from us
Please tell us all what definition you used to say a station was rural? I’ll pull them up directly so folks can see..
I predict you have no answer
I have a dumb question.
I have been reading about UHI for years and still don’t understand what the issue is and why it must be adjusted for.
Is it because it is warmer downtown versus the rural area around the city? I don’t think so, because the anomaly data should take care of that – right?
##############################
Lets just do a simple toy example
A 0 0 0 1 1 1 2 2 2
B 0 0 0 0 0 0 0 0 0
Station A and B start out rural. A develops into a large city
taking anomalies will not remove the trend in A which is
due to UHI . That is the concern
If it because of the baseline – that downtown Minneapolis might have a greater population now than in 1951 or 1961 or whatever, and therefore it is warmer today than 1951 or 1961 and we want to adjust for that?
I guess my question is what is the adjustment and why do it?
If it is warmer downtown Chicago today than at some point in the past or even warmer in downtown Chicago today versus rural today – isn’t that the temperature? Isn’t the instrument reading correctly?
I apologize if this is a stupid question – but I really need a basic answer to explain what the big deal is about UHI.
its not a stupid question. There are a few wrinkles.
See the toy example above. The concern is that the warming
at station A is caused by UHI not by climate change. So
some skeptics have argued that the warming we see in the record is UHI.. that is, they argue that the stations are
located in cities and that cities warm over time due to factors
other than GHGs.
The other wrinkle is something like London.
Let me do another toy example
L: 0 0 0 1 1 1 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3
During the first period when the city is growing we see warming due to UHI.. then the city stops growing ( or its spreads ).. the city core stays the same .. and so the warming due to UHI levels off.
for folks who want a picture of all the GHCN stations daily
have fun
http://gis.ncdc.noaa.gov/maps/ncei/cdo/alltimes
Kenneth Fritsch:
The problem with this idea is Steven Mosher will go on long-winded diatribes about how stupid and lazy and dishonest I am, and he’ll do so while making remarks on the issues that are wrong. His responses to me upthread contain a number of errors on technical matters, but I know better than to expect me pointing the errors out would do any good. If I address only his substantive points, the discussion will degrade to nothing but insults until he just leaves. If I discuss the personal and derogatory remarks, then I’ll just waste everyone’s time. Either way, nobody will find the discussion worthwhile because of the incessant insults and derogatory remarks.
That exact pattern has played out plenty of times in the past. That’s why I no longer have discussions with Mosher. In fact, I basically don’t respond to him at all. I’ll point out some of the falsehoods he writes and highlight the poor nature of his comments, but that’s it. I’m not going to waste my time, or anyone else’s, trying to have a meaningful discussion when one is impossible.
I agree wholeheartedly with this. Implemented properly (people can argue about which cases are), breakpoint analysis is a far better tool than just relying on metadata to try to figure out what non-climatic influences have affected the data and trying to account for them.
I just felt it was important to point out such approaches are not objective, with the implication they can be proven to be correct on some fundamental level. That’s impossible, which is why, as you say, it is important to test the implementations that people create.
Hmmm….
Downloading the monthly adjusted GHCNftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v3/ghcnm.tavg.latest.qca.tar.gz, and then counting unique entries yields:
.
grep 2015TAVG ghcnm.tavg.v3.3.0.20160715.qca.dat | wc -l
.
2650
.
Then if you wanted to be hard ass, but have more reliable stations, exclude those missing one or more months:
.
grep 2015TAVG ghcnm.tavg.v3.3.0.20160715.qca.dat | grep -v “\-9999” | wc -l
.
1555
.
That’s certainly what I found looking at TMAX daily data.
.
The poor distribution, unreliability, fragmentation, and lack of availability of global data preclude making any justifiable statement about global daily TMAX. TAVG doesn’t appear to be much better. But the other data sets ( RAOB, MSU, SST ) have comparable issues. And that’s not to denigrate the efforts of those (BEST, et. al.) charged with making sense of things – that’s what we consumers want. But there’s a lot of fuzz around all the climatology data sets.
“Put another way – if the correction is to adjust for changing population in an area over time – why do we need to adjust for that?”
I will give you a sketch of where the argument is going. Some folks say stupid things about smearing and smoothing.. I think Anthony watts started this argument, others who should know better just mimick him. Now, the idea here is that there are many stations that are urban and when you compare them with rural neighbors the UHI signal is spread into the rural. So one question is Ok how many urban are there?
Any way.
The first thing I want to do is dispel the notion that the network has this huge number of urban stations.
To do that I’m going to show people how many stations are in various types of enviroments. First we will use population because that is what many people glom onto.. but later we will add other variables.
Now one reaction is entirely predictable. When you show people
what 1000 people per km actually looks likes.. they do the following
A) they take conclusions from cities ( with say 1 million people)
and they extrapolate and say, that there should also be UHI
( reduced ) at populations of 1000. Of course there is
not a lot of research to back up this supposition, some
isolated stuff here and there and then they do the following.
B) You have to prove that the lightly populated areas DONT have
UHI or whatever they want to call it. So, You’ll note the shift. The shift in argument from showing UHI in cities… to demanding proof that this doesnt exist in say less populated places.
At various points through these arguments folks will continue to reference studies of UHI on large cities.. all the while scrambling on Google to find a random paragraph here or there in a paper that talks about sub urban UHI… or they just try to change the topic to adjustments or Micro site.
The next step will be to eliminate suburbs from the data.. and just be left with ex urban rural and natural.
At this point.. some will continue to argue that there must also be some bias in these stations.. But again, we see that what started as a concern grounded in observation ( big cities are warmer) becomes a concern grounded in nothing.. it just becomes a “what if” concern.. what if small changes to sites.. cause huge problems..
At that stage we are probably at the question of micro site. And THERE we are back on some firmer observational ground..
So the journey here is from a concern grounded in observation
( big cities have UHI ) then understanding that the data is not mostly big cities… shifting to a concern that suburbs and smaller cities may also suffer from some reduced for of UHI .. ( and we note that the observational grounding gets weaker and weaker ) shifting to a generalized concern about bias or a demand that one prove bias doesnt exist.. few folks are aware of how they are shifting from an observationaly gounded concern to a ‘what about unicorns” concern.
And finally, if folks follow things they can land back in observationaly based concerns.. specifically micro site
Steven Mosher (Comment #149381):
Thank you for your response.
So we are trying to figure out how much warming is caused by CO2 or other GHG’s, and we want to back out UHI and I suppose natural variability.
So we want a handle on UHI so we can back it out.
I get that.
Now, I can see how we could adjust the temperature of downtown Chicago today to adjust for the population growth from a century ago (or whatever baseline period you choose).
But why bother?
Philosophically, whether the warming is due to CO2 or air conditioning, or concrete or asphalt or population growth – it is still anthropogenic – right?
Why not measure the temperature in each location as well as possible, take the data for what it is, and compare it to locations where there are no people and figure out all increased temperatures due to humans, no matter if it is more body heat, more air conditioning, more waste heat, more concrete, other GHG’s, land use changes, aerosols, etc.?
But thank you for your explanation – it does help me understand what the perceived problem is.
Steven Mosher (Comment #149387):
I didn’t see this post until I posted last one.
Thanks again.
Ya. TE
generally speaking I avoid GHCN-M..
Put another way when I have clients who want temperature data ( outside of BE work ) I just use GHCN-D and the other daily data.
The dataset is updated daily.
last I heard they were going to revamp GHCN-M and include more data. we will see.
“Philosophically, whether the warming is due to CO2 or air conditioning, or concrete or asphalt or population growth – it is still anthropogenic – right?”
yes but the concern is that the sampling of stations is skewing the estimates of global warming.
Good questions. Hopefully when we are done.. You’ll be able to see that.
1. UHI is real… Yes the literature shows that large cities ( 1 million or more ) have an artifical bias of say 1-2C
2. That bias doesnt show up in a global record because
A) there are not a lot of large cities in the data relative to everything else.
Lets just do a quick and dirty stupid guesstimate.
Suppose I had 100 stations.
90% are not biased by UHI
10% are in super big cities with a 2C bias for UHI.
When you average it all what kind of final bias will you see?
Now lets do the opposite..
Suppose you THOUGHT that 90% of the stations were urban
and you thought that UHI bias was easly 2C
and then you average these 90% biased with 10% unbiased..
well, you’d think.. hey all the warming is just UHI bias.
later when we get to adjustments we will be able to see how much of the bias can be detected and removed by an algorithm..
In other words.. can an algorithm that knows nothing about the population of city detect that it is warming more than its neigbors and try to correct for that?
Sometimes yes… sometimes no.
Daily TAVG does include a lot more:
.
grep TAVG 2015.csv | cut -d \, -f 1 | sort | uniq | wc -l
.
7037
.
Of which,
.
grep TAVG 2015.csv | grep -v US | cut -d \, -f 1 | sort | uniq | wc -l
.
4693 are outside the US. Not nearly the coverage density as the US, but still a lot more than daily TMAX.
.
I’m guessing the reason is that the TAVG OCONUS sites take the average of hourly obs instead of TMAX/TMIN thermometers.
.
Standardization would be nice. But then, so would living in peace and harmony.
“I’m guessing the reason is that the TAVG OCONUS sites take the average of hourly obs instead of TMAX/TMIN thermometers.”
Err No.
Long ago say 1990 a list of stations was selected to maintain as a discrete SUBSET of all the data
Toy example
10 stations… in 1990, they decide that they want longish records so they select stations 2,4,6,7,9,10
years go by..
in 2000.. stations 9 and 10 are decomissioned and your count goes down
And then in 2000 they start a whole new program like CRN
so you have stations 11,12,13, 14
These new stations dont get added to GHCN-M. they get put in daily.
.
Rick A:
You correctly ask “Why not measure the temperature in each location as well as possible, take the data for what it is, and compare it to locations where there are no people and figure out all increased temperatures due to humans…” The correct answer is that there are insufficient data world-wide at truly anthropogenically-unaffected locations to make such a direct comparison.
What Mosher chronically fails to acknowledge is that outside the contiguous USA, NW Europe, and Australia, virtually all records long enough to provide a reasonable indication of secular trend in the presence of natural multi-decadal oscillations (i.e., preferably >120yrs) are located somewhere in the vicinity of sizable population centers (>50 thousand). There are virtually no historical stations located in truly pristine rural areas; there was no incentive to establish such, nor could they be maintained without daily human attention. The thousands of records constantly trumpeted as “non-urban” by the misleading population-density criterion–typically at airports–are often demonstrably subject to anthropogenic effects of man-made materials and activities, not to mention downwind transport of heat from urban centers.
Contrary to the impression promoted by BEST, the global data base is not at all representative of the actual–overwhelmingly rural–globe when it comes to the low-frequency content of temperature records that determines the “trend.”
BTW, here’s the aggregate difference between the deviations from the 20th- century mean of vetted urban (>50K pop.) and non-urban (<50K pop.) station records geographically representative of the contiguous USA:
http://s1188.photobucket.com/user/skygram/media/Publication1.jpg.html?o=0
Note the suggestion of a logistic curve governing the urban – non-urban discrepancy and the cooling effects of earlier introduction of MMTS instruments at urban stations in the 1990's.
Sky, you are right on. The whole point is that stations were not set up to measure global temperature; they were for local weather use. Their importance was to city populations, agriculture, aviation (after the 1950s) and that is were they were placed and manned. The sea records are worse but off topic.
.
Very few can argue after learning all the inhomogenities in the data that it needs adjusting. The danger is when hundreds of decisions are made about which data to use or weigh bias is inevitable. Normal scientific protocol requires this type of scenario to be subjected to adversarial inquiry. Unfortunately, BEST was supposed to be that adversarial component. In adversarial I mean sincere and transparent investigation coming from opposite null hypotheses.
.
When Steven first claims I am wrong the claim there is suburban contribution to UHI because UHI is “urban heat island,” this is a clue to adversarial discussion.
.
Steven, I have been thinking about whether to start trying to abide by your deals to answer an arbitrary question you pose in exchange for continued productive dialogue. I am thinking that is unfair and may be against the spirit of Lucia’s rules. I think you should just defend your point, concede or remain silent and let the reader decide based on what is written already.
.
Local anthropogenic surface warming is generally in line with degree of development, cultivation and habitation. This happens mostly gradually and mostly non-reversibly throughout the temperature record. It should not be counted as global warming for two reasons:
1) It’s local to the small areas where we are measuring, not global.
2) GHG is the “alarm” not homes, fireplaces or pavement.
“What Mosher chronically fails to acknowledge is that outside the contiguous USA, NW Europe, and Australia, virtually all records long enough to provide a reasonable indication of secular trend in the presence of natural multi-decadal oscillations (i.e., preferably >120yrs) are located somewhere in the vicinity of sizable population centers (>50 thousand). There are virtually no historical stations located in truly pristine rural areas; there was no incentive to establish such, nor could they be maintained without daily human attention. The thousands of records constantly trumpeted as “non-urban†by the misleading population-density criterion–typically at airports–are often demonstrably subject to anthropogenic effects of man-made materials and activities, not to mention downwind transport of heat from urban centers.
Contrary to the impression promoted by BEST, the global data base is not at all representative of the actual–overwhelmingly rural–globe when it comes to the low-frequency content of temperature records that determines the “trend.â€
A good the first person to try to make verifiable claims about
” virtually all records long enough to provide a reasonable indication of secular trend in the presence of natural multi-decadal oscillations (i.e., preferably >120yrs) are located somewhere in the vicinity of sizable population centers (>50 thousand).”
1. First bad assumption. Sky thinks you need long records
to detect a trend. You would think he read the great
skeptics on this.. you dont need long records. Long records
are only needed if you work in ANOMALY land.
2. At least he pulls a number out for population ( 50K)
But note… you need to know the AREA
“There are virtually no historical stations located in truly pristine rural areas; there was no incentive to establish such, nor could they be maintained without daily human attention. The thousands of records constantly trumpeted as “non-urban†by the misleading population-density criterion–typically at airports–are often demonstrably subject to anthropogenic effects of man-made materials and activities, not to mention downwind transport of heat from urban centers.
Note the next two testable claims
1. Non urban are at airports
2. Downwind of urban centers
Now this is funny.
1. I can guarantee you that sky NEVER checked the wind directions from urban to non urban, especially in historical data
2. he cannot cite any literature that argues UHI is advected over kilometers and kilometers.. forgetting that heat rises
3. he seems to think that every site is collocated with man made surfaces.
So as I predicted.. what started as a theory based in observations about urban centers becomes un supported claims about all non urban stations being downwind of urban
because the wind only blows in one direction.. and rural sites are all downwind..
too funny.
So again.. as predicted. we go from observationally GROUNDED concerns about UHI in cities… to speculation
A) UHI can be transported horizontally for great distances
B) all non urban sites are at airports
C) There are no pristine sites.. in other words.. there are micro site issues
Its probably better sky if you just say… ya its all micro site..
“Sky, you are right on. The whole point is that stations were not set up to measure global temperature; they were for local weather use. Their importance was to city populations, agriculture, aviation (after the 1950s) and that is were they were placed and manned. The sea records are worse but off topic.”
So tell me Ron..
For stations in 1900.. where are they?
whats the population?
Big cities in 1900?
more than 1000 people per sq km?
Where exactly were they all placed?
Do you see how you have transitioned from an observation based argument ( look at the UHI in Seoul ) to speculations about site locations?
So, tell us more about where these sites were located.. perhaps the airports in 1900.. opps.. maybe the car traffic then.. opps.
How about Lewis and clarks records..
or colonial forts…
Suddenly we are not talking about measured UHI… suddenly you are asserting that there had to be some kind of UHI.. back then… because.. well there had to be..
But cool.
Now we can go back in time..
care to make predictions..?
Good question for sky.
How far downwind does the UHI extend?
5 km from the urban core?
10km?
In the megacities field studies what did they find?
A city of 1 million will see a UHI on the order of 1-2C at the core. what will this value be 10km from the city core?
is the UHI “spread” uniformly?
if wind speeds of 7m/ses kill the UHI in a city.. what then?
Since cities increase the rainfall downwind in some locations
( typically less than 20km) are they actually cooling downwind locations? what does the research show
What differences are there between canopy layer UHI and boundary layer? why is that important?
what about the presence of greenspaces as we move out of the city.. how do they impact the cooling of the UHI?
perhaps the berlin study can help you.
Steven Mosher (Comment #149393)
July 15th, 2016 at 1:49 pm
“later when we get to adjustments we will be able to see how much of the bias can be detected and removed by an algorithm..
In other words.. can an algorithm that knows nothing about the population of city detect that it is warming more than its neigbors and try to correct for that?
Sometimes yes… sometimes no.”
From this comment I would surmise there is an independent procedure in testing the adjustment algorithm – at least with regards to suspected UHI. This interests me. Could we fast forward to the punchline on this one?
Ron.
When Steven first claims I am wrong the claim there is suburban contribution to UHI because UHI is “urban heat island,†this is a clue to adversarial discussion.
Err no. I am trying to keep you precise. You see, as I have
pointed out, when you talk about UHI in cities, you are on
FIRM observational grounds. you cant point to many studies
they all measure in the city and then in the rural areas.
When you ASSERT that there is a sub urban contribution.. you
are ASSUMING and not working from observational studies.
That is why you had to go looking for suburban and why you
only found a dissertation on SUHI.. did you not think I was one
step ahead of you.. ( the RPA stuff is good surprised you read
a 248 page dissertation and missed it )
IF there is a suburban “UHI” that will have to be established.
you havent. you’ve just assumed it. What was your null?
how did you test it?
.
“Steven, I have been thinking about whether to start trying to abide by your deals to answer an arbitrary question you pose in exchange for continued productive dialogue. I am thinking that is unfair and may be against the spirit of Lucia’s rules. I think you should just defend your point, concede or remain silent and let the reader decide based on what is written already.”
its pretty simple. you have asserted, with no support in data and no support in literature that there must be a sub urban form
of UHI. I have only refused to accept your assertion.
It should be pretty simple. You believe there has to be UHI in the
record because you based your belief on studies of cities.. you cite
them all the time.. Next, when you find out that 90% of the stations are not in cities… what can you do? Well, you cant pull out the studies on suburban UHI.. because there are very few..
So you have to switch from an observation based position to
a “well there has to be UHI in suburbs”
So lets take the next step…
Lets elimate the suburbs.. you cited London as an example where UHI extended into the suburbs.. and Sky also thinks UHI blows downwind..
Lets eliminate the suburbs.. we do that next.
.
“Local anthropogenic surface warming is generally in line with degree of development, cultivation and habitation. This happens mostly gradually and mostly non-reversibly throughout the temperature record. ”
1. generally inline? where did you get that from?
2. Degree of development. what does that mean?
here is a question.
A) UHI , the literature from Oke finds, that a city of 1Million
will have an average UHI of 1-2C. lets call it 2C
B) if we have 1/2 of the development.. do we have 1/2 of the UHI? in your mind? in the literature?
3. You call it irreversible, but then you cited literature that showed otherwise.. what gives?
In any case we are making some progress.. and in a little bit
we can look at all of Sky’s claims, and we can apply the taxonomy of RPA. its in the 248 page dissertation you read and referred to..
It should not be counted as global warming for two reasons:
1) It’s local to the small areas where we are measuring, not global.
2) GHG is the “alarm†not homes, fireplaces or pavement.
“From this comment I would surmise there is an independent procedure in testing the adjustment algorithm – at least with regards to suspected UHI. This interests me. Could we fast forward to the punchline on this one?”
It is probably premature. but eventually I will get there as I am usung this excerise to build up a repository of stuff that just runs
TE exactly.
The population of temperature stations is amazingly small.
25 in the Antarctic which means 1 in a hundred of all weather stations in GHCN is in the Antarctic which is perhaps 1/6th of the earth land surface
2475 for the rest of the earth .
Which is about 2 for every big urban city and 1 for each airport and 1 for each state in each country in the world.
And they only give data for 5.9 years on average before being replaced
Why people have to exaggerate the number of active stations to make it look like they are doing something is beyond me.
43k fiddlededee sticks.
“The population of temperature stations is amazingly small.
25 in the Antarctic which means 1 in a hundred of all weather stations in GHCN is in the Antarctic which is perhaps 1/6th of the earth land surface
2475 for the rest of the earth .”
Wrong.
do you read?
GISS, BE, and CRU all use SCAR
NCDC does not,
NDCD only uses GHCN-M
but since GHCN has so few stations in antarctica.. other folks
all supplement GHCN with SCAR
from GISS:
“Graphs and tables are updated around the middle of every month using current data files from NOAA GHCN v3 (meteorological stations), ERSST v4 (ocean areas), and SCAR (Antarctic stations), combined as described in our December 2010 publication (Hansen et al. 2010). These updated files incorporate reports for the previous month and also late reports and corrections for earlier months.”
Total stations I have in antarctica is 89.
‘2475 for the rest of the earth .”
Wrong again.
SM 149380
I use Richard Mueller’s criteria to define a station as urban or rural .
I predict that you will find this OK.
An urban station is one in a city where the population is greater than 50,000.
A rural station is any station where the population is less than 50,000 ie non urban.
NB
“Urban areas are heavily overrepresented in the siting of temperature stations: less than 1% of the globe is urban but 27% of the Global Historical Climatology Network Monthly (GHCN-M) stations are located in cities with a population greater than 50,000. If the typical urban station exhibited urban heating of the magnitude of Tokyo, this could result in a severe warming bias in global averages using urban stations.”
R Mueller.
Please do not pull up the stations but instead you could address Professor Mueller’s assertion which seemingly contradicts some of your assertions.
Mind you I still agree with you as to how irrelevant it “should” be.
SM 149409
Only one of us is playing games with numbers here.
When you descend to sophistry to win you lose.
Let’s try to agree on numbers.
43k you said it, not me and left it unspecified.
It is the basis of your essay yet you never specified which data set you are using.
Which data set are you using?
BEST GCHN or not?
If you wish to use what “other folks” use instead of what you used, say so at the start.
If you wish to use SCAR then put up the number of stations and the data set incorporating scar in the 43k.
“SM 149380
I use Richard Mueller’s criteria to define a station as urban or rural .
I predict that you will find this OK.
An urban station is one in a city where the population is greater than 50,000.
A rural station is any station where the population is less than 50,000 ie non urban.”
Sorry. That is not his criteria.
That is the intruduction of the paper where he talks about OTHER PEOPLES SOURCES…
Finish reading the paper
!e analysis presented here is based on merged monthly average
temperatures from the Berkeley Earth Surface Temperature Study
dataset. !is dataset consists of measurements from 36, 869 unique
stations, which are merged from 15 preexisting data archives (the
dataset and a description of the merging and $ltering can be found at,
http://berkeleyearth.org/dataset/). We classify these stations as rural
or non-rural by comparing their locations with the MODIS 500m
Global Urban Extent classi$cation map (MOD500) of Schneider
[24,25]. Schneider used Collection 5 MODIS 500-m resolution satellite
imagery to classify land use as urban using supervised decision trees,
a statistical learning algorithm that they trained using a set of sites
with known land cover type. !ey de$ne urban areas to be “places
that are dominated by the built environmentâ€. Urban heat islands
are primarily a result of replacing the natural (soil, vegetation, etc.)
surface of the land with buildings and arti$cial ground surfaces, which
makes the MOD500 dataset potentially quite helpful in identifying
built-up regions that may be subject to urban heating. It may provide
a criterion that is less socio-economically biased than night lights
data; therefore , it o”ers an alternative to the approach used by GISS.
!e MOD500 map is available as a raster image, providing a binary
classi$cation (urban or not urban) for a global grid with pixels of size
15 arc-seconds. According to Potere the MOD500 map outperforms
other global urban maps in terms of predicting city size and per pixel
agreement on a sample of known cities with population greater than
100,000 [26].
SM 149409
I said GHCN had 25 stations in the Antarctic note only 3 inland at elevation and mueller says extremely difficult to assess variation in altitude of temperature.
You said I was wrong but then said “GHCN has so few stations in the Antarctic that they need to be supplemented”
Did you mean they have less than 25 stations?
GISS, BE, CRU NCDC
Which Data set are you using?
Or do you just wish to use all of them to give “I have 89 stations” as a throwaway line?
If you can pick 89 stations out of 4 data sets when challenged how many stations overall would you claim there are ?
Certainly not 43k
–
I repeat Richard Muellers BE temperature averaging process said from a high of 5883 GHCN stations in 1969 to a precipitate decrease to 2500 in 2013.
Take 25 GHCN stations in the Antarctic off (you did not dispute the GHCN) figure gives 2475. My maths is right.
–
Finally Richard says 44,840 effective stations over several hundred years average station life 5.9 years.
BE GHCN is as close to 43k as I can get from your potential range of data sets.
If you wish to use BE total or CRU or NCDC which all incorporate GHCN and SCAR etc you must end up with a lot more than 43k which means you should modify your initial essay.
And put in the data set,please.
And I still think you are doing a good job putting the concept up.
My maths is right
angech
43K stations. as you know thats the number in our dataset.
and yes SCAR is one of the sources we use.
149413
That is not his criteria?
It is my definition of urban and rural based on population numbers.
It certainly is his criteria.
Influence of Urban Heating On the Global Average Temprature Land Average using Rural Sites Identified from MODIS Classifications 2013.
The quotes on Urban areas being <1% of the globe but 27% of the stations being located in cities with a population of 50,000 were cut and pasted.
His comment, his criteria, my definition.
Muller not Mueller, my bad.
All OK .
As you said there are many ways to define population and being urban or rural. No one definition is yet correct. That is what your post is about!
149381
Does it matter?
Sort of.
Why?
I guess because some people use change in land surface temperature measurements to argue the case of CO2 induced global warming.
Without considering that other factors can be at play.
Like areas of UHI.
Some people would like the effect of UHI removed so the temperature rise due to CO2 can be more accurately assessed or ascribed.
As you say it is a storm in a teacup.
Although Muller disagrees.
Steven Mosher writes this in response to a user:
I don’t know if that particular user can “cite any literature that argues UHI is advected over kilometers and kilometers,” but I know other people can. Not only do I personally know of studies which have argued an effect is present at 10 kilomters or more on average (meaning some days will be higher, others lower) for some urban areas, I’ve even read a study arguing heat from London’s Urban Heat Island can be advected kilometers downwind.
I’m not commenting on the validity of anything anyone is saying here, and my silence shouldn’t be taken as indicating anything. That remark just happened to catch my eye while I was skimming through the comments, and I thought it might be useful for people to have some idea what distance outside a city UHI can have an effect.
It gets more interesting when you start looking at how UHI impacts precipitation.
Ah brandon your spoiling all the fun… make them find it.
and 10km… for some days, for some cities, and for some wind speeds
how does roughness factor play a role?
“That is not his criteria?
It is my definition of urban and rural based on population numbers.
It certainly is his criteria.”
Nope not his criteria.. do you know who actually wrote that?
so sky and ron.. if you want to stick with observations ( and not modelling of UHI ) then there is a new lab ( hint– one of Oke’s old associates ) that does have some data.. that may or may not help you.. be careful.. read carefully….
If you cant find it.. I’ll point to it later
and.. be careful about classifying..
( one of the rural sites is actually industrial.. and well, 23 of the 27 are within the city limits and very well populated– rural fraction as measured by NVDI is something different .. )
Steven Mosher, your post is written as a serious scientific essay to address the known interests of the bloggers on this site. Yet when your assertions are skewered you seem to become litigious about exacting definitions. When your point still does not hold water then you seem to say “it was all in good fun.”
.
You seem to be saying you know that your essay is misleading but it’s purpose is to provide a healthy educational exercise to motivate skeptics to do more research confirming that many scientifically written essays are knowingly misleading. And, to show goodwill about it you promised you would have “helped us” is we did not find the conflicting facts on our own. I do not know if this fits in well with the BEST intentions as specified in your non-profit tax filings and on your about us page:
.
Steven, I still very much appreciate your information and especially that we get it gratus. I am just suggesting to motivate our research by telling us what there is to find in advance of us finding it.
Ron,
I’ve told you a bunch of times to read more. The path we will take here is up to the questions that get asked and the assertions that get made. Like advection and what counts as Urban, and the relationship between SUHI and UHI.
So, Im putting you in the place of the scientist actually having to make decisions.
Let’s take your reference to SHUI.. without studying it you just randomly search for suburban UHI and find a dissertation.
You dont look at the research about the relationship between SUHI and UHI you just link. Had you looked you would have found the LA studies and the Birminham studies at the BUCL. And you would probably wouldnt have linked to the dissertation..
Let’s take Sky’s reference to advection. Up until recently that has usually been studied with models (Like the London modelling study that found effects up to 40km)
More recently you would have found the BUCL ( if you followed Sue Grimmond you would have known about this earlier )
That would have led you to the urban network around Birmingham, 20+ sensors and also temperature sensors ( about 100) on lampposts.. and their advection study.
Then you could make smarter arguments.
Like I said, based on your responses we can go in different directions.. There are no final results to post.. its being written as you guys make suggestions– there are of course some paths I have gone down before.. and some I havent..
here is the map to an advection study.
Note the locations of the sensors within the urban boundary
http://www.nature.com/articles/sdata201638/figures/1
The point will be simple.. we want to make sure that we dont
pick sites that are close to urban centers.. or that we would
classify as “rural” just because of the NVDI figures..
To save you the trouble Ron, I have added some maps
to show you the BUCL study area.
Paradise location is taken ( in the study) to be the city center
with a LCZ of compact hi rise.
The other sites around this ( within 15km or so ) are assigned a
“rural” score of sorts depending on their greenness
W04 and W07 represent the extremes of the stations within the
study area..
Steven, thank you for your continued prompt replies.
.
When you say: “I’ve told you a bunch of times to read more,” I hope you realize that I am not retired but and work full-time as most other skeptics do. I was drawn into these blogs by reading a WSJ article by Christy and Curry about 2 years ago. Before that I was a big believer in global warming and sea level rise. If I had not looked for myself I would have been giving a standing ovation to a professor’s commencement speech 7 weeks ago claiming that we have just 30 years before “climate launch” (when today’s record highs will be the average temp.,) and seas will rise by 9 feet by the end of the century. I was the only one sitting; even my wife and children stood up and applauded along with the crowd and new graduates. I’m sure Ken Rice’s and Josh Halpern’s speeches on climate to our best and brightest are similar.
.
BEST’s charter is to inform the public. I realize there is no contract holding you here or anywhere else. You are acting above and beyond your employment duty surely. But for the 200K of public money salaried to Elizabeth Muller I am hoping you impress her with the need to post things like detailed descriptions of how your process and climate index’s are crafted. Specifically, I would be interested to see links to info and graphics of:
1) Global map of station locations in 30-yr snapshots for entire record.
2) Typical instrumentation and sampling methods globally on timeline.
3) History of adjustments and evolution of methodology.
4) Honest UHI assessment. Your aerial imagery above would be more informative to the subject is you included aerial snapshots of Las Vegas metro every ten years since 1950. Or, refer to Hamdi’s Brussels study.
5) Have a SKS- type section citing all the areas of current true consensus and which topics are under debate.
6) On the topics under debate provide the best argument from each side. Perhaps invite guests from each side to update periodically.
.
The reason SUHI is critical, as you know, it goes right to the heart of the BEST method of looking for breaks. For there to be a clear break between urban and suburban stations there must be assumed no suburban UHI. The further you get from urban the more pristine the station but the break is muddied due to the less expected correlation from distance.
.
On the topic of statistical break method, days of high UHI are likely going to occur in a city and surrounding suburbs simultaneously thus lifted the baseline and dampening the break.
.
Finally, UHI per se introduces no bias to the global climate record. That is agreed. It the change in UHI gradually over 165 years that biases the record. So moving stations from cities to parks and airports did not solve the problem. It worsened it; because most of the warming effect was done before the move only to be repeated again in the new interval.
.
The only way to adjust properly for UHI is to quantify it with 30+ year comparison studies of urban to suburban, suburban to rural and rural to pristine untouched area stations if they exist. Then add each station a down adjustment based on an LCZ (Oke) evaluation. If the adjustment could be made accounting for weather conditions like wind, clouds and rain then there should be no breaks with neighbors on any day. At this point I would be satisfied as a scientist that you have cleansed the data from local AGW.
So is your current position:
SUHI is uncertain or SUHI is negligible(both locally and for global average)?
Mike. Suhi is different than uhi.
And advection plays a role.
So the next step will be to eliminate stations that are close to cities. .
Ron suhi is different than uhi. And not very well related.
As for your requests… This is all on my time. I will get around to discuss in a bit. At lunch
Mosher says: “Sky thinks you need long records
to detect a trend. You would think he read the great
skeptics on this.. you dont need long records. Long records
are only needed if you work in ANOMALY land.”
This claim is on par with his notion that getting the log of zero is “simple.” Oblivious to the frequency response of linear regression, he doesn’t even begin to distinguish between truly secular trends (implied by AGW) and the slopes of irregular, natural multi-decadal cycles.
Mosher then attributes to me the following:
“1. Non urban are at airports
2. Downwind of urban centers
Now this is funny.
1. I can guarantee you that sky NEVER checked the wind directions from urban to non urban, especially in historical data
2. he cannot cite any literature that argues UHI is advected over kilometers and kilometers.. forgetting that heat rises
3. he seems to think that every site is collocated with man made surfaces.”
What I actually wrote was: “[O]utside the contiguous USA, NW Europe, and Australia, virtually all records long enough to provide a reasonable indication of secular trend in the presence of natural multi-decadal oscillations (i.e., preferably >120yrs) are located somewhere in the vicinity of sizable population centers (>50 thousand). There are virtually no historical stations located in truly pristine rural areas; there was no incentive to establish such, nor could they be maintained without daily human attention. The thousands of records constantly trumpeted as “non-urban†by the misleading population-density criterion–typically at airports–are often demonstrably subject to anthropogenic effects of man-made materials and activities, not to mention downwind transport of heat from urban centers.”
Never mind that Mosher has no concept of my professional work. Those capable of reading closely will note disturbing differences between these verbatim quotes. Such presumptuous, straw-man argumentation doesn’t deserve any further response.
OK, so you will come back to that later.
Your answers look rather Stokesian, but I will withhold judgment until you get back to that.
Is it fair to say UHI is .1-.2C based on 10% of 1-2C, or do adjustments deal with that?
MikeN: “Is it fair to say UHI is .1-.2C based on 10% of 1-2C, or do adjustments deal with that?”
.
I just read Peterson 1999 on rural stations (assumed rural zero delta UHI for populations <20K) versus the entire GHCN (including those rural). Result: although the warming in the rural group of 2290 stations was 0.80C/100yrs and the whole 7280 GHCN was 0.92C/100yrs they tracked together identically until the last year, 1998. Peterson attributed this to poor data supply on last year for rural group. Conclusion: Although there is UHI and it's dangerous to public health it does not affect stations over a hundred-year interval at all. Move along, nothing to see.
.
Then came Hansen et al(2001) who found 0.1C/100yrs UHI in periurban (suburban?). So what gives? Parker(2010) explains it this way:
.
But Hansen, using extrapolation from urban UHI to periurban sites is convinced there is zero rural UHI. Well — that may because rural UHI is either microsite or land use LULCC, which Hansen considers AGW and not needed to be adjusted for. Steven M promises to talk about that later.
.
Who was wrong, Steven, Peterson or Hansen? What was the flaw?
Can anyone tell me if this post got changed? I don’t see any indication it’s been updated or edited, but I swear it seems rather different than before. Is that just my imagination?
Ron
You still owe me a few answers, trust me, as you answer questions or make assertions it helps to shape where this thing is going..
I have a sense of where I want to take it, but there are many many pathways.. think of it as a new way of doing blog posts.. I know
that you and others want me to make a claim so that you can play skeptic. But I want to try something different..Sky, for example, was super helpful by bring up certain topics that I wanted to cover so.. I will go a bit down the paths he was suggesting.
But to get to your questions
links to info and graphics of:
1) Global map of station locations in 30-yr snapshots for entire record.
That is coming. one of the issues is fitting this all in memory
I may have to switch to my azure system and well that costs
me personally. But it should be doable.. its in my “plan”
2) Typical instrumentation and sampling methods globally on timeline.
Not even sure what you mean by this or how it relates to the
question of UHI and population.. definately back burner
3) History of adjustments and evolution of methodology.
The best bet there would be SVN. work in progress
dead ends, all that is probably in the chief scientists
SVN. I dont have access to that. only when folks want to
publish something do they move it to the public version.
4) Honest UHI assessment. Your aerial imagery above would be more informative to the subject is you included aerial snapshots of Las Vegas metro every ten years since 1950. Or, refer to Hamdi’s Brussels study.
A) The imagery was meant to save you time and embarassment. I hope it succeeded.
b) The goal here is to exclude Urban sites. LV is pretty
dang urban, but since you asked I will try to show
you what we know about it going back in time.
c) Brussels study is cool, I’ll look for a logical way to weave that in, but since that site is going to get classified as urban
its not all that interesting.
5) Have a SKS- type section citing all the areas of current true consensus and which topics are under debate.
Cool. send money. Our current focus is on air pollution and
nuclear. You can read a piece I worked on about the consensus
its one our site.
6) On the topics under debate provide the best argument from each side. Perhaps invite guests from each side to update periodically.
Nice idea. No funding for that.
###################################
.
The reason SUHI is critical, as you know, it goes right to the heart of the BEST method of looking for breaks. For there to be a clear break between urban and suburban stations there must be assumed no suburban UHI. The further you get from urban the more pristine the station but the break is muddied due to the less expected correlation from distance.
SUHI is surface uhi.
uhi in suburban doesnt actually go to the heart of break points. breakpoints have more to do with discontinuites.
That said, the concern about being CLOSE TO an urban center
is one that I share. Basically, long ago when I did my first
“urban” catagorizer ( basically a filter that tries to divide urban
from non urban ) the issue of being close to urban popped up
its ugly head. And also the issue of green parks within urban.
( now you can understand why I put up the picture of central
park ) we have to be careful when looking at a site in isolation
it may have zero population, but be covered in pavement
( hint.. Long ago I found this with airports.. and industrial zones.. ) and then folks may have reasonable concerns about being adjcent to urban centers. Observationally we only have a
couple studies that look at this. A few SUHI studies but those
are problematic.. a few modelling studies.. But the BUCL stuff is really cool. And we will walk down that road for a little bit
.
On the topic of statistical break method, days of high UHI are likely going to occur in a city and surrounding suburbs simultaneously thus lifted the baseline and dampening the break.
Ah well, the problem is this. take the BUCL study for example.
1) they had to focus on cloud free days and anti cyclonic days
to find UHI in the urban core. So it happens infrequently.
2. When it did happen the core was +4C and the areas
surrounding the core ranged from +.5c to +1.5c
I dont see how one can flatly proclaim that this will
destroy breaks.. In then end.. if you have one station
with a +X UHI and the suburban neighbors with 1/4X.
the field should relax to the 1/4x number.
.
“Finally, UHI per se introduces no bias to the global climate record. That is agreed. It the change in UHI gradually over 165 years that biases the record. So moving stations from cities to parks and airports did not solve the problem. It worsened it; because most of the warming effect was done before the move only to be repeated again in the new interval.”
1. It is interesting that so many people think station moves
happen a lot. and when they think that the moves are
all similar.
2. Your analysis depends on the notion that UHI happened
gradually over time.. So we can actually go look at some
of these questions.. questions about airports and moving
from cities to parks.. and see what kind of numbers there.
Any way.
I added a few charts..
I’ll summerize later
.
The only way to adjust properly for UHI is to quantify it with 30+ year comparison studies of urban to suburban, suburban to rural and rural to pristine untouched area stations if they exist. Then add each station a down adjustment based on an LCZ (Oke) evaluation. If the adjustment could be made accounting for weather conditions like wind, clouds and rain then there should be no breaks with neighbors on any day. At this point I would be satisfied as a scientist that you have cleansed the data from local AGW.
MikeN
‘Is it fair to say UHI is .1-.2C based on 10% of 1-2C, or do adjustments deal with that?’
I think I wasnt clear. i was giving you a hypothetical.
“Can anyone tell me if this post got changed? I don’t see any indication it’s been updated or edited, but I swear it seems rather different than before. Is that just my imagination?”
As folks raise questions or issues, I am putting up new charts at the bottom as I explore there concerns.
So, Sky raised the issue of advection.. So, I put up some charts of the BUCL area and a couple views of what the area looks like so we can have a common set of stuff to discuss.
So, as folks raise issues and concerns or questions that I can answer I will add stuff to the end of the post.
Like: what happens to our station count if we only looks at stations that are not urban and not adjacent to urban, using BUCL as a guide.
Next post.
For the next post, may take some time, I will be doing this.
1. Every site will be given a population classification. This is
somewhat arbitrary but traceable to the RPA classification
in the dissertation that Ron referenced. There are 6 classes
Natural, Rural, Exurban, suburban sprawl, dense suburban,
and urban core. They are based on densities. we will still
look at raw density, but when drawing maps its easier to
understand and visualize if we bin the data. of course we
can use other binning.
2. We’ll also be careful not to ignore adjacency. There are two
ways to do this. one is quick the other is not. The quick way
is this. From every site we evaluate the neighborhood.
If there is any densely populated grid square.. we call that neigbhorhood urban.
3. I’ll do Rons charts for stations over time. 30 year bins
and we will count up the population types. Then we can
also look at the changes from start of the period to the end of the period. This will actually be a first.. so who knows what we will find..
149423
“Nope, not his criteria do you know who actually wrote that?”
I gave the full name of the paper.
He is one of the leading authors.
Since you ask who wrote it you must concur that criteria are in the paper.
Since his name is on the paper, prominently, he endorses those criteria in that paper, at that time.
–
“SUHI is surface UHI”
Is it? Or is it surburban UHI , I am a little confused here, nothing new.
–
I feel I have also missed which data set Steven is using in this essay.
The one with the 43 k stations. 149415 comment.
Does anyone know and would they mind posting it here?
–
UHI means urban heat island.
Suburban does not come into it, or rural, semi rural or very rural or industrial.
Some of these categories are quasi population based.
Some of these areas will obviously have heat gradients relative to the urban areas.
The issue is one of urban heat islands.
Steven mentions problems of once having defined urban
e.g. 50,000 people in a city, that not all urban areas have the same characteristics or even necessarily form a heat island.
That is tough.
You have your definition and you go with it.
You can always break it up and look into it further but a city and a population mass 50,000 are reasonable criteria to use.
Non urban heat island and gradients are another issue, related but not part of the thrust of this discussion.
–
Measuring the temperature when you have only 2500 stations, 27 percent of them in cities (Muller), and a limited station lifespan 5.9 years average (Muller) to try to track changes over makes estimates in the change of the temperature very difficult.
One eminent scientist stated (Muller was an author on the paper , he endorsed but did not actually write it) “If the typical urban station exhibited urban heating of the magnitude of Tokyo, this could result in a severe warming bias in global averages”) that UHI could give a severe warming bias, I know, cherry picked to extremes.
Okay, yeah. I’m definitely not interested in this thread now. Steven Mosher explains:
Maybe I’m alone in this, but I’ve always thought making substantial changes to posts without indicating the changes is bad. It’s a shame as I think UHI is a fascinating topic.
angech:
It’s surface, not suburban.
Hamdi(2011) compares the temp trend in Brussels as compared with a nearby rural station for 50yrs (1955-2006) in summer months using both ground station records and remote sensing IR analysis to diagnose a temperature trend independently to demonstrate an unbiased method for confirming UHI at particular defined locations by their surrounding IR reflective profiles.
.
Hamdi found:——–aerial diagnosis ——ground observed
50yr (urban-rural)—- 0.8C —————– 0.9C
Tmax increase——– 0.06C/dec ———–0.06C/dec
Tmin increase——– 0.15C/dec ———–0.19C/dec
.
Fujibe(2011) Diagnoses UHI for Tokyo as 0.2C/dec over a 100-yr period, which corresponds to Hamdi.
.
McKitrick and Michaels(2007) used a method studying urban vs. non urban temp trends to find a 0.13C/dec difference in land temps (0.30C recorded -0.17C corrected for urban artifacts). That attributes a whopping 43% of the land record.
.
Ron,
Confirmation bias much?
I wouldn’t cite M&M2007 as evidence of anything. It has major flaws. Mosher could go into more detail, but I doubt it’s worth it.
MosherBias. I just coined a new Climate Science Term.
Andrew
DeWitt, if you don’t want to rely on Mosher as your only information you can see the detailed story of the MM07, the Gavin Schmidt09 attack and the McKitrick and Nieremberg(2010) rebuttal here and here on CA.
Ron
Mckitricks errors in 2007 are hilarious.
Among the problems.
1. He spread the population of countries equally over all the grid cells.
2. he assumed an equal growth rate in population for every grid cell.
The first mistake ends up putting huge populations in antarctic– because there are grid cells there where the country listed in ghcn is france and england. The same for St Helena island.
The first mistake also gives you a population in the Gobi desert that is equal to every other grid cell in china. And yes Alaska gets the same treament in population as Chicago,
One of the things I wanted to do was to redo his analysis with the right data for population. some of data he referenced was not available.. but I dont know.. you know he used literacy as a regresssor for temperature? yup..
The other thing he failed to do in his regression was separate latitude from population. ( they are somewhat co-linear) and approach he used on regressing latitude out wasnt correct either.
I’ll put up another post in a bit.. its sunday…
Here Ron
http://www.mdpi.com/2072-4292/8/2/153
“Figure 6a shows large differences between LST and Tair data collected at the time of the satellite overpass. These differences vary with landuse (Table 5) and range from around 3 °C in suburban areas, to over 13 °C directly adjacent to the thermal core, further highlighting the significance of the different processes contributing to UHIsurface and UHIcanopy in these areas. For comparison, an intensive study of Los Angeles, using 44 meteorological stations and seven AVHRR images during three days in August 1984, indicates a 5.4 °C difference between radiant surface and air temperatures in the afternoon (standard deviation of 2.3 °C) [28].
“Direct comparisons between Tair and LST at night, show that Tair is consistently higher than LST across the city, ranging from 0.7 to 3.2 °C (Figure 7a). Temperature differences again vary with landuse (Table 6) with the lowest temperatures differences between LST and Tair in the city centre, likely because of the increased thermal capacity of urbanised surfaces. In contrast, the largest differences are in areas with more vegetation (i.e., Sutton Park and Woodgate Valley Country Park).”
#####################
LST is cool like I said I spent a lot of time looking at it.
1. to try to understand the spread of UHI beyond the core
2. to try to estimate UHI for small cities
3. to try to estimate SAT for areas where we dont have data in the arctic.
The problem was you could find places where it worked..
and then other places where it didnt
Next, it only works on clear sky days ( when UHI is going to be higher )
But if you like SUHI then read Peng
SUHI for 419 large cities.
SM: “Mckitricks errors in 2007 are hilarious”
.
Steven, he was claiming only that the data shows urban warming. He used proxies for urbanity not for precision but for qualitative analysis.
.
Even if Dr. McKitrick et al. were flat wrong it wouldn’t justify the flagrant breach of ethics shown by Schmidt and journal editor of IJOC, putting Schmidt in charge of the review process for McKitrick’s rebuttal to Schmidt. It’s already a Davey and Goliath contest; why does Goliath also need to be the referee? Only because power has no shame to “stop the leaks.”
.
SM:
.
McKitrick:
“McKitrick:
Fewer than one-third of the weather stations operating in the 1970s remain in operation.[2010]”
Again just untrue, and ignorant. He is talking about GHCN Monthly, which consists essentially of two parts. One is a collection of historic records, made by a grant funded project in early 1990’s. The other, continuing from then, is a subset of those stations which report monthly under the CLIMAT system, starting about 20 years ago. CLIMAT places demands on the met stations and processors, so the list was rationalised. Most of the original stations remain in operation, and their records are in GHCN Daily, or ISTI, for those that want them.
NS: “Again just untrue, and ignorant.”
.
I realize that BEST utilizes GHCN Daily. Does GISTEMP or CRUTEM? Was McKitrick’s statement ignorant in 2010?
.
BTW, good to see you, Nick.
.
Everyone, Nick has done an awesome job at producing interactive tools at his web site and blog at https://moyhu.blogspot.com/ It’s purely climate oriented. I applaud that.
.
I was not paid by anyone to say that.
Ron,
Thanks. But
“Was McKitrick’s statement ignorant in 2010?”
Yes, it was. It claimed that only one third of stations remained in operation. That has no basis, and it is ignorant because it refuses to look at what GHCN Monthly actually is. It is a particular list maintained for rapid and reliable monthly use – so indices can get their averages out within a fortnight of so of end month. That is what CLIMAT achieves. The fact that some stations don’t keep up with that reporting schedule doesn’t mean they aren’t operating. And yes, GISS and NOAA use GHCN M, but CRUTEM uses a different set, and I don’t think they made a similar change in early ’90s.
Of course, the homogenisation of GHCN does, or at least can, use other data on a more relaxed reporting schedule.
ps I wrote about Ross’s paper at the time in 2010 here, explaining what was wrong with it, and how and why the changes from historic GHCN going forward were made.
Why is reporting data in the age of the internet such a burden? It seems like little there’s been relatively lax protocols to make reliable records all the way back and to this day. I believe this is a frustration that Dr. McKitrick and others share.
.
Forget rigor, it just makes common sense to have protocols that control random error and systematic bias.
.
It’s also common sense not to allow adjustments decades after the observations based on conjectures and one team’s bias. If mistakes are made time is of the essence to correct. For example, if I notice that I have been short shipped from my vendor a month after receiving the parts or wait a month before notifying them that is a different situation then if the error is found upon receipt and notification is made immediately. The reason is as time passes the number of possibilities for motivating bias grows at the same time evidence is fades.
1) There was no excuse to allow stations to choose their time of daily Max/Min observation or to allow them to change it at whim.
2) The method for ship buckets for measuring SST should have been tested for reproducability the day it was thought of. The bucket used should have had specifications as well as the thermometer and time for stabilization.
3) When switching SST measure from engine intakes there should also have been bucket tests done concurrently to establish the re-calibration at the time before discarding the bucket observations.
4) Specific times of day and sea depths needed to be specified. I have found neither.
5) All methods should have had built in periodic statistical quality checks on going to ensure operators stayed true to original methods.
6) If an oversight is found years later to have been made independent teams need to do field testing and compare results to establish an official correction to be accepted internationally.
.
Having Dr. Karl or Dr. Thompson decide arbitrarily one day there needs to be more warming, and it’s time to get to work on it, just makes a mockery out of climate science and all science.
.
That the answer to the UHI problem was simply to relocate stations to areas not yet affected as much and call it a day seem astounding to me. I understand that CRUTEM does not even make a UHI adjustment. I see papers by Peterson(1997) proving UHI has no impact (zero) on the temperature record and Steven Mosher (sometimes) making the same claim in 2016.
.
Nick, Steven, I hope you realize the stakes involved are much more then the governmental actions you desire at the moment.
“Steven, he was claiming only that the data shows urban warming. He used proxies for urbanity not for precision but for qualitative analysis.
.
Even if Dr. McKitrick et al. were flat wrong it wouldn’t justify the flagrant breach of ethics shown by Schmidt and journal editor of IJOC, putting Schmidt in charge of the review process for McKitrick’s rebuttal to Schmidt. It’s already a Davey and Goliath contest; why does Goliath also need to be the referee? Only because power has no shame to “stop the leaks.â€
you meant Dr. Schmidt.
“He used proxies for urbanity not for precision but for qualitative analysis.”
look if you put millions of people living in antarctica and millions living on st Helena, and in the gobi desert, and never drew a map to check ( which was how I found it ) you cant even get the qualitative correct.
second it’s funny how you drop your interest in precision.
finally I have no problem having opponents even enemies review my work.. I’m answering your questions arent I?
Gavin could have done a better job since the problems are basically data selection and processing errors.
Nick Stokes (Comment #149483)
“Fewer than one-third of the weather stations operating in the 1970s remain in operation.[2010″
He is talking about GHCN Monthly, which consists essentially of two parts. One is a collection of historic records, made by a grant funded project in early 1990’s. The other, continuing from then, is a subset of those stations which report monthly”
–
Untrue and ignorant?
Hilarious ?
–
Well since you guys are undoubtably intelligent throwing mud in a tag team shows concern which in turn suggests veracity for McKitricks comment.
It also reflects badly on your character that you resort to the tactics you accuse others of.
–
Best to point out that the statement refers to the number of weather stations, not compilations as Nick would have one believe.
The statement that less than a third of the stations operating in the 1970’s remain in operation is completely true.
The average life span of a GHCN station is currently 5.9 years, Nick.
Think about the maths a little.
GHCN splices station changes Nick.
As Mosher is so keen to say, it’s a new station Nick.
By his standards and definition you would be lucky to have 5% original stations left.
McKitrick is being kind by considering spliced stations to be original, and even in that case he is right. Station dropout and non replacement is the biggest fudge factor in global temperature assessment going.
No wonder you have to sit on the lid and denigrate people.
Angech,
I’ve posted the stations over time and maps over time as Ron requested. on a new post.
Now, the funny thing is you keep making assertions without knowing the facts
“The statement that less than a third of the stations operating in the 1970’s remain in operation is completely true.”
False.
You fail to distinguish between a station being in operation
and
A station stopping its monthly reporting TO CLIMAT.
again just use daily data because some stations still in operation may stop reporting monthly figures for GHCN-M to ingest
any way you can see the number of stations on the latest post.
I’ll be back later I have to report june.. ~19K stations.
u and ross too funny
http://berkeleyearth.org/temperature-reports/june-2016/
Here is the “fresh” results from probably the largest US investiagtion on UHI, Thomas Karl for the period 1901-1984, i believe around 400 city-rural pairs.
http://hidethedecline.eu/media/city%20heat%20IPCC/aau.jpg
This was published just before the climate debate/science according to some got less scientific, thus the label “fresh”.
If anyone do not agree with this old analysis, please tell where Karl did it wrong.
Some station are so UHI infected that hardly anyone denies it, ex. in USA Los Ageles, Boston, New York etc. What interesting is that HadCRU use such data in connection with so called climate science. Would not hurt to skip such stations. Or like BEST that use Copenhagen , Budapest, etc in stead of the many many other long stations available.
“Why is reporting data in the age of the internet such a burden? It seems like little there’s been relatively lax protocols to make reliable records all the way back and to this day. I believe this is a frustration that Dr. McKitrick and others share.
.
He is not frustrated at all about it. Neither are you. internet has nothing to do with it. CLIMAT son
https://www.wmo.int/pages/prog/www/OSY/Publications/TD1188/HandbookCLIMAT-CLIMATTEMP_en.pdf
Now, If I were NOAA I would just deprecate USHCN and GHCM-M
Ron
“It’s also common sense not to allow adjustments decades after the observations based on conjectures and one team’s bias. If mistakes are made time is of the essence to correct. For example, if I notice that I have been short shipped from my vendor a month after receiving the parts or wait a month before notifying them that is a different situation then if the error is found upon receipt and notification is made immediately. The reason is as time passes the number of possibilities for motivating bias grows at the same time evidence is fades.
1) There was no excuse to allow stations to choose their time of daily Max/Min observation or to allow them to change it at whim.”
##################
pretty much illustrates your ignorance of the history.
read more. comment less and make fewer mistakes.
Data collection in the US was for a long time largely done by Volunteers.. unpaid volunteers.
Go look at the us coverage.
Next
“”It’s also common sense not to allow adjustments decades after the observations based on conjectures and one team’s bias”
A) its common sense to correct mistakes you find. Go ask Leif about his work on centuries old data and correcting that.
B) There are multiple teams working indpendently on the same problem.
Let me give you an example.
http://www.homogenisation.org/files/private/WG1/Bibliography/Method_Description/Climate/pandzic%20etal.pdf
canada also does its own adjustments
Austrailia… many countries.
And yes the USA does its own country series.
So you have
Source Decks
Global adjusted Inventories
Individual Country adjusted Inventories
CRU
CRU draws from the Individual country adjusted repos.
So for example they use the 200 station version of canada
adjusted by canadian experts.
NDCD
NCDC draws from GHCN-M and they do their own
AUTOMATED adjustment process.
GISS
GISS draws from NCDC adjusted and they add their own
adjustments
BE
we draw from source decks and do our own automated
adjustment process.
So you have multiple teams drawing from different but overlapping data, using different methods ( Expert adjusted, data driven, etc ) all coming to essentially the same answer.
You seem to be trying to give the impression that there is an issue with this. I don’t think there is one.
Andrew
2) The method for ship buckets for measuring SST should have been tested for reproducability the day it was thought of. The bucket used should have had specifications as well as the thermometer and time for stabilization.
back in the 1700s? we are lucky that the british had a fetish
for collecting data everywhere with no clear USE in mind.
3) When switching SST measure from engine intakes there should also have been bucket tests done concurrently to establish the re-calibration at the time before discarding the bucket observations.
AH yes, tell all those shipping people to keep throwing buckets over board. Go look through ICOAADS. You wont.
4) Specific times of day and sea depths needed to be specified. I have found neither.
Go look through ICOAADS. A ships intake is where it is.. Look at icoaads. Height of the ship helps in figuring out the depth.
Time of Day? Well there is the proof I need that you actually havent looked. LOOK.. you already have claimed you dont have time to do this job right.. FOCUS.. focus on SST or Satellites.. or UHI… or OHC. FOCUS and make fewer mistakes. Pick one area and dive deep.
5) All methods should have had built in periodic statistical quality checks on going to ensure operators stayed true to original methods.
Too funny. Observational science ( just take sun spots as an example ) have to handle all manner of lapses, changes, shifts.
its detective work.. its NOT lab science. But yes, back when galileo drew images of sun spots he should have been told how to do it correctly.. and since he did not do it up to Rons standards.. its all meaningless junk. throw it out.. all of it.
6) If an oversight is found years later to have been made independent teams need to do field testing and compare results to establish an official correction to be accepted internationally.
fricking globalist.
Data collection in the US was for a long time largely done by Volunteers.. unpaid volunteers.
You seem to be trying to give the impression that there is an issue with this. I don’t think there is one.
Andrew
##########################
Good then you DISAGREE with Ron who claims there is a problem.
Here is what Ron wrote
“1) There was no excuse to allow stations to choose their time of daily Max/Min observation or to allow them to change it at whim.â€
So, go fight with Ron
you wont.
ask yourself why.
I’m addressing what YOU commented. If I wanted to take it up with Ron, I would do that.
Andrew
“I’m addressing what YOU commented. If I wanted to take it up with Ron, I would do that.
Andrew”
Ron: They should not let people change TOB
SM: They were volunteers
Andrew: You seem to be trying to give the impression that there is an issue with this. I don’t think there is one.
############
your world is not what it seems.
A: X is a problem
B: oh, they were volunteers
C: Hey there mr B, you got a problem with volunteers.
B: fact: they were volunteers, ask A why he has a problem
C: No ! I think you should stop going around citing facts.
There is no problem.
B: check with A, he thinks there is a problem.
C: No.. I have no problem with A
B: well you think there is no problem and A thinks there is
a problem and I think there is no problem, so how do
you have a problem with me when we both think there is no problem, but A thinks there is a problem.
C: Did I claim to be logical mr smarty pants.
B: ah no you did not. thank you
So please explain how volunteering causes the problem.
Andrew
The volunteers did exactly what they were told and I have the utmost gratitude toward them. Protocols come from above as do periodic quality checks. It seems that most of the bureaucrats running the show had no appreciation of science and had the exact attitude you voiced: [blank… had a fetish for collecting data everywhere with no clear use in mind.
.
If you were a drug company trying to present your data to the FDA and pleaded your case as above they would put down their pen, lean forward while sliding their reading glasses down and simply stare at you.
.
Steven Mosher of 8 years ago laughed at Hansen’s UHI adjustments, making 50% of them negative to cancel the positive ones out without ever making that clear to anyone.
SM: (2008)
Ron,
“1) There was no excuse to allow stations to choose their time of daily Max/Min observation or to allow them to change it at whim.
2) The method for ship buckets for measuring SST should have been tested for reproducability the day it was thought of”
Who are you talking to here? Who do you think was in charge of all this?
Dr Karl and colleagues didn’t (couldn’t) prescribe how data was collected. They, like you and me, are trying to make sense of the world and how it works, using what data they can get. It’s no use saying, someone should have done this. You have to deal with what was done – what is available.
That applies to GHCN and stations. GHCN doesn’t open and close stations. It compiles, and varies what it includes.
”
Steven Mosher of 8 years ago laughed at Hansen’s UHI adjustments, making 50% of them negative to cancel the positive ones out without ever making that clear to anyone.
SM: (2008) ”
Yes. and so I requested his code so I could do a better job.
I did not berate him with questions. I asked for his code so I could “fix” it.
Turns out… in his next paper he tested one of the things I was sure was a problem.
I was wrong
Steven, did you ever complete your inland desert station Mosh index to gauge against cities? I would think that was a good plan if you also remember that deserts would be the most sensitive to GHG due to their low relative humidity and clear sky. But that is exactly the signal we would want to find then adjust proportionately as a control to populated areas.
.
Your 2008 idea of using already “uhi saturated” cities I think was mistaken as can be seen in Frank Lasner’s link above of Karl’s UHI analysis (pre-political Karl). I think you would agree that your yesterday LST paper link analyzing canopy from also shows the issue grows in proportion to area. I would be hard to find stations in any areas that have not changed in 100 yrs or even 50 except rust belt and dead coal towns.
You see Ron you have a choice.
As you argue you dont have time to read the science.
My suggestion: focus. on the science.
SM: “I was wrong.”
.
Don’t sell yourself short. He had a huge advantage over you.
.
So explain how 50% of the adjustments were positive and 50% negative and how there were double the number of both types in the USA versus the rest of the world?
.
I don’t mean to berate. We are genuinely curious.
‘The volunteers did exactly what they were told and I have the utmost gratitude toward them. ”
too funny.
First you say they cannot change TOB at whim.. Now you say OK..
hilarious.
Okay, lets try another pass at this.
.
How did Peterson(1997, 2003) not find any UHIE when comparing rural to whole set?
.
a) The trend for UHIE is present at most stations, regardless of classification.
b) There is no UHIE.
Steven Mosher, do you think the recent SST adjustments that Ron Graf discusses above are valid?
Ron,
Let me give you an example of how a volunteer organization works. I’m a licensed SCCA Flag & Comm worker. I don’t get paid. I get the occasional T-shirt plus food and beer on Saturday night. We’re supposed to be the eyes and ears for the stewards, reporting what’s happening on the course as well as displaying flags to the drivers and being first responders for crashes.
A few years back, the stewards in one geographic region decided that there were too many instances where cars hit each other without causing a spin or accident. So they told us to report on the radio and produce written reports on any instance where we thought there had been contact between two or more cars.
A few races later, we were told that the program had been a success and that there were now very few incidents being reported. Guess what. The number of incidents hadn’t actually changed, we just stopped reporting them because it was a PITA to write an incident report.
If you get too officious with volunteers, they either ignore you or quit.
Frank Lanser:
Karl’s un-politicized results on UHI in the USA are very much in line with what I found not by pairwise comparisons, but by constructing geographically representative estimates of the areal average temperature for the lower 48 states based upon exclusively urban and non-urban records, each of whom covered the entire period 1896-2005. The ~0.7C rise of the urban – non-urban discrepancy would correspond to Karl’s UHI for a city of roughly half a million.
Your remark “Would not hurt to skip such [UHI-corrupted urban] stations” would be spot on, were it not for the fact that in most of the continents there are precious few non-urban, intact, century-long stations whose records could provide the 20th-century backbone of the historical record. In fact, I cannot even find fifty such records in all of Asia, Africa and S. America–combined. All of the methods of obtaining “global average temperature” are simply ad hoc schemes for covering these glaring lacunae in geographic coverage. So much for “settled climate science.”
“Steven Mosher, do you think the recent SST adjustments that Ron Graf discusses above are valid?”
So in a post about population what have I heard
1. There are only 1500 active stations
2. Ross Mckittricks paper
3 Too many sites by cities of 50K or more and by airports
4. What about TOBS
5. Lets look at SST
And people wonder why more scientists dont come on the web to answer questions or engage with the public.
WRT to SST
Long ago ( See climate audit ) I started to look at ICOAADS. Steve Mc started to look as well. Why? because if you want to say something worth while you actually have to look at at data
Do you thnk angech would persist in his folly if he actually went to the Berkeley site, download the data and used the software I would give to him to actually see for himself? No he would not persist in his folly if he actually did some work.
Continuing, after spending a bunch of time looking at the data organizing it, seeing what data fields were there and not there I hit a beautiful brick wall: memory size. the point? I’m not an expert.
Later ( last year or so ) we started a project to re do the oceans using a method that is very similar to a approach sugggested by Mcintyre. Again its an approach where you dont do individual adjustments.. I have the math around here somewhere ( basically minimizing a set of equations ) there was one road block…
How you handled NMAT could give you two different answers.. our
“discovered” adjustments ended up being close to the discrete approach..
There are another couple papers in the works so I’ll hold off commenting
So
karls adjustments LOWER estimates for ECS. Do I think they are valid? Some work we did justified them under certain assumptions.
Second there is more work coming out on Karl.. so wait ( skeptics wont like it )
I dont know how to convince skeptics that if they want to be taken seriously they should go look at actual problem areas..
like aersols and clouds..
Also. there is an opportunity to do some good work on SST.. picking at karl isnt the opportunity given what I have seen..
“So explain how 50% of the adjustments were positive and 50% negative and how there were double the number of both types in the USA versus the rest of the world?
.
I don’t mean to berate. We are genuinely curious.”
Go do your OWN DAMN work on hansen. Look I spent a good deal of time pouring over docs, downloadn=ing files, begging for code, getting the code, struggling to get it to run, looking at every case,
every line of code.. and it was all pretty much useless as far as
ADVANCING OUR UNDERSTANDING.
part of the problem was hansen using nightlights to identify rural.
nightlights is not a good proxy.. So in the US you might have
places with no people but a bright light close to the site. In other parts of the world you have tons of people but no lights
Plus he originally used the wrong file for nightlights so I sent an email to the guy who posted the file and had him deprecate the file and ( as I recall ) they switched to the better file.. After he switched and did his “pitch black” sensitivity tests I stopped paying attention to what he was doing ( 2010 ) WHY? because I dont care about improving his approach.. I wanted to do my own and learn from his mistakes.. In the end.. a new approach is only marginally better– worse if you ask someone like brandon.. whatever..
So you want to know aboout hansen? go do the damn work or pay me for the years I wasted and I will tell you what I remember
So that’s a yes, and you think the attacks against SST adjustment are likely illegitimate?
I think you had lots of responses about your original topic.
To understand “why more scientists don’t come on the web to answer questions or engage with the public” one need only look at the patently self-promoting polemics by scientifically incompetent blog lions.
Steven, judging by your comment information to wordcount ratio I would swear you are spamming your own post.
.
Ron Graf:
DeWitt: “…how a volunteer organization works.”
.
I have been an organizational volunteer most of my adult life. SM’s assumption that I didn’t know the US COOP stations where run by volunteers was a canard since he had me refresh him in the whole TOBs spiel 3 weeks ago, my point was there was zero discipline in quality control in recording the data by administrators, NOT volunteers, who certainly would have continued the evening observations, not disturbing the control. NOAA /NCDC was at fault for being oblivious for over a generation.
.
Steven, how many times has NOAA, NASA or Hadely instituted a new method or code based purely on theory without synthetic data testing? On CA I see this over an over a McIntyre criticism.
.
Bottom line — Poor data combined with poor analysis quality control, combined with political opportunity, is certain disaster. Because if your right, not enough people will believe you and if your wrong nobody will believe anyone the next time.
.
Was Peterson right or wrong? This is your post.
I volunteer at a pregnancy center. We collect data from clients. The data that is easy to collect we get. The data that is hard to get we don’t always get.
The issue with “volunteers” is part of the AGW narrative that is designed to distract from and confuse the real problem, which is that a reliable system/process was never established then to get the information that is desired now. The information doesn’t exist.
So you can either approach this problem honestly, or you can make things up that fit your political narrative.
Andrew
Nick Stokes:
“Having Dr. Karl ….”
Not to nitpick, but Tom Karl has an honorary doctorate. I actually give him more credit for his accomplishments attained without an academic doctorate.
Ron,
Weather stations were instituted for weather observations. They were never designed for climate change studies. They work fine for weather, where the bias from TOBS is insignificant, not so much for long term change. In metrology terms, they are not fit-for-purpose. But, they’re all we have. Whinging about quality control in the past is, IMO, pointless.
Thank you Steve for putting up the data set that you used, Berkely Earth Temperatures. I was using BE GHCN data which was the best I could find on google searching after reading your post.
I did find the BE stuff with some struggle after reading your post and have left my biased responses there.
It would have helped if you had put up a link, still would actually because not everyone reading and commentating here is up to speed on the references. Could you possibly append it above and on your second post, the actual data set, to avoid this problem.
Thanking you in anticipation.
I was under the impression that if there’s no QC present, there’s no science present, whatever the date. But I guess the days of serious science and scientists are over.
Andrew
Whining….
.
One might be led to believe by Nick and Steve that climate change was not even a concern before 1980 and thus poor quality control of data should be forgiven. And that the need for strict quality in particular to have ability to separate out UHI from the climate was not considered. Contrary to this, I am finding that concerns of effects of UHI on the climate trend as well as station moves and instrumentation changes have been voiced all through the 20th century. Here is a quote from Oke(1972):
.
Karl(1988) acknowledges concerns about microsite (though he does not use that term), yet it took Anthony Watts in the 2000s to expose it and embarrass the establishment to do something about it.
.
The detachment of scientists from the stations almost reminds me of an anthropologist’s approach of savoring over the garbage left strewn at an earlier time, except in this case the inspection period (over 50 years) has been concurrent with the remnants.