Guest post by David R.B. Stockwell PhD
I read with interest GHCN’s Dodgy Adjustments In Iceland by Paul Homewood on distortions of the mean temperature plots for Stykkisholmur, a small town in the west of Iceland by GHCN homogenization adjustments.
The validity of the homogenization process is also being challenged in a talk I am giving shortly in Sydney, at the annual conference of the Australian Environment Foundation on the 30th of October 2012, based on a manuscript uploaded to the viXra archive, called “Is Temperature or the Temperature Record Rising?”
The proposition is that commonly used homogenization techniques are circular — a logical fallacy in which “the reasoner begins with what he or she is trying to end up with.” Results derived from a circularity are essentially just restatements of the assumptions. Because the assumption is not tested, the conclusion (in this case the global temperature record) is not supported.
I present a number of arguments to support this view.
First, a little proof. If S is the target temperature series, and R is the regional climatology, then most algorithms that detect abrupt shifts in the mean level of temperature readings, also known as inhomogeneities, come down to testing for changes in the difference between R and S, i.e. D=S-R. The homogenization of S, or H(S), is the adjustment of S by the magnitude of the change in the difference series D.
When this homogenization process is written out as an equation, it is clear that homogenization of S is simply the replacement of S with the regional climatology R.
H(S) = S-D = S-(S-R) = R
While homogenization algorithms do not apply D to S exactly, they do apply the shifts in baseline to S, and so coerce the trend in S to the trend in the regional climatology.
The coercion to the regional trend is strongest in series that differ most from the regional trend, and happens irrespective of any contrary evidence. That is why “the reasoner ends up with what they began with”.
Second, I show bad adjustments like Stykkisholmur, from the Riverina region of Australia. This area has good, long temperature records, and has also been heavily irrigated, and so might be expected to show less warming than other areas. With a nonhomogenized method called AWAP, a surface fit of temperature trend last century shows cooling in the Riverina (circle on map 1. below). A surface fit with the recently-developed, homogenized, ACORN temperature network (2.) shows warming in the same region!
Below are the raw minimum temperature records for four towns in the Riverina (in blue). The temperatures are largely constant or falling over the last century, as are their neighbors (in gray). The red line tracks the adjustments in the homogenized dataset, some over a degree, that have coerced the cooling trend in these towns to warming.
It is not doubted that raw data contains errors. But independent estimates of the false alarm rate (FARs) using simulated data show regional homogenization methods can exceed 50%, an unacceptable high rate that far exceeds the generally accepted 5% or 1% errors rates typically accepted in scientific methods. Homogenization techniques are adding more errors than they remove.
The problem of latent circularity is a theme I developed on the hockey-stick, in Reconstruction of past climate using series with red noise. The flaw common to the hockey-stick and homogenization is “data peeking” which produces high rates of false positives, thus generating the desired result with implausibly high levels of significance.
Data peeking allows one to delete the data you need to achieve significance, use random noise proxies to produce a hockey-stick shape, or in the case of homogenization, adjust a deviant target series into the overall trend.
To avoid the pitfall of circularity, I would think the determination of adjustments would need to be completely independent of the larger trends, which would rule out most commonly used homogenization methods. The adjustments would also need to be far fewer, and individually significant, as errors no larger than noise cannot be detected reliably.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
We have done a shadow post of David’s paper on the NCTCS blog:
http://theclimatescepticsparty.blogspot.com.au/2012/10/is-temperature-or-temeperature-record.html
Anyone interested in attending the AEF conference mentioned in the second paragraph please note that David will be speaking on the 20th (Twentieth) and NOT the 30th as mentioned above
Victor Venema: In your response to laterite you acknowledge that your objections to his argument are not valid. If you want to rebut it, why not explain why reference homogenization is NOT a logical circularity instead of picking nits?
“AndyG55 says:
October 15, 2012 at 1:54 pm
@Victor “The period of urbanization is typically in the order of 30 years. After this there is no longer a bias in the trend, the temperature just has a fixed constant bias, which does not affect trend estimates.”
And this is EXACTLY why temperatures have leveled off for the last 15 or so years. Thanks you !!!”
Got it in one AndyG55.
There is so much wrong with the logic in “Victor’s” statement I am not sure he has actually ever sat down and looked at the blatant unproved assumptions he is making.
The only bit of it that is right is that urbanization induces a bias in recorded temperatures.
Assumption 1: Once a site is urbanized that there is no change in rate of bias from changing urban use.
Assumption 2: There is no change in the rate of site urbanization (i.e. the number of sites being urbanized).
The fact is that you can only claim a fixed constant bias from urbanizationfor a single site as long as the area of urbanization around the site remains constant.
Neither Assumption stands up to even cursory scrutiny.
UHI “adjustment” is often performed by lowering the early temperature records in adjacent rural areas. Don’t ask me to justify this bizarre procedure – I can’t. The result of this approach is to produce a greater linear trend after the UHI adjustment than before, and that alone should tell anyone with half a brain that a negative sign has been dropped.
While I’ve given up expecting these temperature adjustments to make much sense, is it unreasonable of me to expect that at least the nonsense should be consistent? So is the “irrigation cooling effect” (ICE?) then dealt with by raising early temperatures in adjacent non-irrigated areas producing a lowered trend? It seems not! There is no consistency here. It is all ad hoc.
… an ad hoc-key stick …
Victor Venema
So you are going down fighting. I admire that.
Victor Venema
It’s not for me to say, but I think you are welcomed here any time.
You did cause a stir, even though it’s been homogenized.
Re: Tom G(ologist) comments. Bingo! Another reason to eliminate the EPA. Their original intended job is done. When is Congress going to do anything about it? (ans.:never)
Victor
I am not a scientist so I have to rely on what makes sense to me.
You said ………” And why the trend in the rural stations is about the same as the ones in the urban stations.” Are you saying they should be increasing at the same rate because of AGW?
That seems wrong to me because if rural areas are basically unchanged and something makes the temperature go up there could be AGW (or maybe some other GW). My understanding is urban areas are unique types of heat sinks. I would expect urban areas to have temperature increases based on the rate of urbanization and build a warm bias. It would seem to me growing urban areas should not be used to identifying AGW.
Although, I would guess Dr. Hansen has an adjustment number handy.
Am I missing something or am I just wrong?
I’m one of the skeptics who believes we could still be exiting the LIA. I would not be surprised to see another 2 degrees C warming before the next big ice age. Not much evidence for runaway CAGW but lots of evidence of ice ages. I don’t believe the loss of the Arctic Ice cap would end the world. Losing the glaciers in Ohio didn’t hurt. They farmed Greenland. The ice caps have melted before. This time, a little warming with more freed-up land and water will help feed and sustain mankind. Some of us think that’s a good thing.
cn
Circular reasoning is the mark of a religious belief.The classic god said, therefore its true. Its true because god said.This being one of the reasons society adopted the scientific method.Its first principle is to slow down the bloodletting of fixed convictions by the acceptance of the concept that I/we might be wrong, therefore let us reason using testable methods. Is it possible that we have cycles in mass human nature and its down into unreason we go again?
The principal weather site for a region is chosen to be the best in terms of quality and duration and non-missing data. If an algorithm flags that a heterogeneity is present, then selects a nearby site or sites for homogenisation, there is no prior evidence that the latter sites have qualities that will improve the principal site. They have errors of their own; often it is impossible to recognise these errors because of lack of reporting of meta data etc.
So, what is the point of selecting a primary site then probably making it worse (because you cannot usually know that you will make it better) by comparison with nearby sites. Should you not first compare the nearby sites for differences from the principal site, then correct them for heterogeneities before feeding them back as adjustments? This is a form of circularity that it would be hard to express more clearly. It is merely smearing the errors.
I take exception to the publication of correlation coefficients like those from Victor Venema that “For these five years, the r2 between the 60 monthly USCRN and USHCN version 2 anomalies is 0.998 and 0.996.” (Menne et al. On the reliability of the U.S. surface temperature record. J Geophys. Res., VOL. 115, D11108, doi:10.1029/2009JD013094, 2010.)” The correlation coefficient will vary with length of data (= number of observations), whether it is Pearson or rank or another variant. It will be different for daily data over the same period as averaged monthly data and different for averaged annual data. There is an easy way to test this. Take a single weather site, lag its daily data by 1 day and then try the variables I’ve listed above, original against lagged. You’ll probably also find that Tmax has a different R to Tmin and other interesting confusions. In other words, climate studies benefit from a statistician’s input from the very start.
I simply do not believe that any major tempertaure/time series has it right so far for the last century.
There is no point in sophisticated further analysis like the calibration of proxies until the fundamental temperature/time series can pass tests of reliability.
This is one reason why there is so much junk in climate papers. It’s a golden rule to set the foundations firmly before you start building.
One of the reason, I give SkS better grades for science, is reading comprehension. The comment of Peter S above is a good example. There is a disturbing discrepancy between what I wrote and what Peter seem to think I wrote.
But actually, I was comparing the scientific quality of the posts at WUWT and SkS, not the comments. Unfortunately, the posts at WUWT show the same lack of reading comprehension, at best. For example the post on the conference abstract by the Greek hydrologists, who falsely wrote that the temperature trend is between 100% and 50% of the reported value, which Anthony even more wrongly reported as being 50%. Whether this is a problem with reading comprehension or a lie, I leave up to the reader.
J. Philip Peterson says: “Victor Venema It’s not for me to say, but I think you are welcomed here any time.”
Thank you. I must say, reading many of the above comments, I do not feel welcome, but I had not expected much different. I know WUWT and thus know that there is no argument that will ever convince the locals.
Chuck Nolan says: “Are you saying they [urban temperatures] should be increasing at the same rate because of AGW? That seems wrong to me because if rural areas are basically unchanged and something makes the temperature go up there could be AGW (or maybe some other GW). My understanding is urban areas are unique types of heat sinks. I would expect urban areas to have temperature increases based on the rate of urbanization and build a warm bias. It would seem to me growing urban areas should not be used to identifying AGW.”
Urbanization can gradually increase the local temperature observed at an urban station. If this gradual increase is not representative for the area around the station, this would cause a bias in the estimates of the large-scale temperature trend. Thus if you are interested in this large-scale temperature, you have to remove this local effect by homogenization. If you are interested in the urban climate, you should keep it.
Another option would be to remove urban areas from the dataset. As far as I know, this has also been done and the trend similar. (As I mentioned before, I do know a little about homogenization, but an no expert for urbanization. Bad willing people call this “hand waving”, I prefer not to lie about my expertise.) The disadvantage of this approach is that you have to assume that your information on the urban/rural character of the stations is perfect and that you remove more data as needed as only a period of the data will typically be affected by urbanization and the rest of the data is useful and helpful in reducing the errors. All in all, homogenizing the data is more reliable and accurate as removing part of the data.
Chuck, I hope that that answers your question.
Geoff Sherrington: “Should you not first compare the nearby sites for differences from the principal site, then correct them for heterogeneities before feeding them back as adjustments? This is a form of circularity that it would be hard to express more clearly. It is merely smearing the errors.”
Yes, you should homogenize all stations simultaneously. I think the simple graphical examples on my blog on the fundamentals of relative homogenization show that this is not circular reasoning.
Geoff, with which quality would you be satisfied? Please remember that every measurement in every science has an uncertainty. The remaining uncertainty in the trends after homogenization have been studied, also in a article of mine, and found to be smaller as the temperature trend observed. That is how science works, you show that your data is good enough to answer your question. Refusing to analyse data and draw conclusions because the data is not perfect is unreasonable and one of the signs of denialism.
It is interesting that the discussion keep on focussing on urbanization, which is not the topic of this post and on which I am no expert as I have admitted multiple times, but that the other problems with this and previous posts are mainly ignored by the locals.
The second figure is particularly interesting. These adjustments are typical of homogenization process (correction of mostly downward jumps). This feature is indicative of anthropogenic disturbance and has been well described in Hansen et al. 2001 (http://pubs.giss.nasa.gov/docs/2001/2001_Hansen_etal.pdf) :
“…if the discontinuities in the temperature record have a predominance of downward jumps over upward jumps, the adjustments may introduce a false warming…”
We can consider this as evidence that most stations are disturbed by urbanization. It is also the demonstration that we should not homogenize the data when we want to use them to evaluate long-term trends.
Victor Venema:
At October 16, 2012 at 2:19 am you assert
NO! That is NOT “how science works”: it is a description of pseudoscience.
No data is ever “perfect”. So, in a real science “you” determine that data emulates reality with adequate reliability, accuracy and precision to provide sufficient confidence that the data is adequate for conclusions to be drawn from it.
It is pseudoscience in its purest form to claim that imperfections in the data should be ignored if the data can provide a desired “answer”.
Therefore, it is necessary for the researcher to provide evidence that the data he/she uses has the necessary reliability, accuracy and precision for his/her conclusions to be valid. In the case of data homogenisation that has not been done. Indeed, the different research teams who provide the various global (and hemispheric) temperature data sets use different homogenisation methods and do not publish evaluations of the different effects of their different methods.
In the above article David Stockwell provides several pieces of evidence which demonstrate that GHCN homogenisation completely alters the empirical data in some cases such that the sign of temperature change is reversed; e.g. compare his figures numbered 1 and 2. That altered data is then used as input to determine a value of global temperature.
It is up to those who conduct the homogenisation to demonstrate that such alterations improve the reliability, accuracy and precision of the data. Claims that such alterations should be taken on trust are a rejection of science.
It seems you do not know the difference between science and pseudoscience so I will spell it out for you.
Science attempts to seek the closest possible approximation to ‘truth’ by attempting to find evidence which disproves an existing understanding in part or in whole and amends the understanding in the light of the evidence.
Pseudoscience decides something is ‘true’ then seeks evidence which supports that understanding while ignoring (usually with excuses) evidence which refutes it.
Richard
V. Venema: But actually, I was comparing the scientific quality of the posts at WUWT and SkS
Can someone please purchase a clue for this fine gentleman?
Victor Venema says:
October 16, 2012 at 2:19 am
“One of the reason, I give SkS better grades for science, is reading comprehension.”
______________
A huge problem of SkS is that their message is controlled. Most here would never be allowed to post there. There are many papers which profoundly influence the debate which will never be discussed or even mentioned- sometimes for weeks,if at all, until it seems, someone comes up with a twist of logic which seems to refute the paper, but which is nevertheless, artfully twisted. At SkS, there is no discussion of scientific topics outside of the party line, so to speak. They rationalize their exclusivity and controlled message and you may ascribe to the rationalization- you would know- but they are violating the most basic tenets of the scientific method. Another issue with that site is that they use and allow techniques better suited to propagandists than scientists, such as the continual use of highly inflammatory terms with which they describe any and all who disagree with them.
If those two issues alone do not send up red flags for you, then there is nothing else to say to you- you either have the consciousness to “get it”, or you don’t.
__________
VV says:
J. Philip Peterson says: “Victor Venema It’s not for me to say, but I think you are welcomed here any time.”
“Thank you. I must say, reading many of the above comments, I do not feel welcome, but I had not expected much different. I know WUWT and thus know that there is no argument that will ever convince the locals.”
_______________
There are others who have expressed to you the same sentiments as J.P.P., in this thread.
Per one of my previous points, the same could not be said of SkS, where any counter point would never see the light of day. You realize that, don’t you?
Your words highlighted above are a blatant attack on the readers, here at WUWT. You have made them over and over, but here you still are. If you do not feel as welcome as you think you should, then remember that what goes around comes around. If you do not believe that or understand it, then again, there is nothing more to say to you about that topic-“let those who have the ears, hear”.
_______________
VV says:
“I am a scientist”.
_____________
We get it.
If we didn’t get it the first time you said it, we surely did the second time.
One gets the feeling that you are using your declaration in order to add credence to your words… employing a logical fallacy, as it were.
There are many here with far greater skills than I posses. As for me, I’m merely an engineer by training and not a scientist. I can read the papers and do the math… I can keep up. . As such, I must say that your own lack of reading comprehension and faulty logic leaves much to be desired; a case in point:
You insist that urbanization (UHI) and station siting have nothing to do with the discussion. Instead, they are really central to the discussion- why homogenize the data, otherwise?
Victor says:
The period of urbanization is typically in the order of 30 years. After this there is no longer a bias in the trend, the temperature just has a fixed constant bias, which does not affect trend estimates.
Ridiculous. The “period” of urbanization for almost any site other than truly rural is as long as the whole record or longer to varying degrees. Urbanization hasn’t stopped or reversed anyway I know of in the US or elsewhere. A few exceptions would be insignificant.
Luther Wu says: “VV says: “I am a scientist”.”
I am, but I never wrote the sentence your quote. How are the reader supposed to trust you, if you cannot even cite right?
Luther Wu says: “You insist that urbanization (UHI) and station siting have nothing to do with the discussion. Instead, they are really central to the discussion- why homogenize the data, otherwise?”
The post is not about homogenization. If you would be interested in learning more, you would pick my brain about homogenization, that is a topic where I can give qualified answers. It would be better to discus urbanization with some one more knowledgeable. If the atmosphere here would be more welcoming these other experts would be more willing to answer.
There are so many more sources of inhomogeneities in climate data, there has been so much economic and technical development in the last centuries. However, somehow the “sceptics” blogs act as if urbanization is the only problem. One get’s the impression that there is a pattern here and that urbanization is a favoured topic because it is the main inhomogeneity that leads to an artificial warming, most other inhomogeneities lead to an artificial cooling of recent temperatures relative to past ones.
beng says: “Ridiculous. The “period” of urbanization for almost any site other than truly rural is as long as the whole record or longer to varying degrees. Urbanization hasn’t stopped or reversed anyway I know of in the US or elsewhere. A few exceptions would be insignificant.”
Urbanization of the complete city can continue for most cities, but the relevant urbanization is the one of the region around the station. The climate trend in the centre of London and Vienna is about the same as the rural trend outside the city. The temperature inside the city is higher due to the UHI effect, but apparently, the effect is no longer getting stronger.
Victor,
You can certainly make the case that what I spoke to you about was not an exact quote.
What I said was a representation of what you said. You know that.
I think you would have enjoyed the ‘how many angels on the head of a pin’ thing from earlier times.
”Citing right’ has nothing to do with the substance of what was said.
“How are (sic) the reader supposed to trust you, if you cannot even…” address the substance of the discussion?
Victor Venema says:
October 16, 2012 at 7:50 am
“It would be better to discus urbanization with some one more knowledgeable. If the atmosphere here would be more welcoming these other experts would be more willing to answer.”
__________________
Do you mean, “more welcoming” as in the manner that we would be welcome at that place to which you keep referring?
Or, what did you mean?
Was your purpose just to launch yet another verbal attack at us?
Those “other experts” can speak for themselves, n’est-ce pas?
Can their assessments withstand the real scrutiny which they would undergo, here?
laterite (David Stockwell),
I read with great interest your discussion of finding an illogical circularity in the commonly used homogenization techniques of time series data on surface temperature.
I am considering what kind of incorrect premises and subconscious biases could explain how both NOAA (GHCN)and NASA (GISS) could embrace a process like homogenization with its logical circularity faults.
Their apparent uncritical use of the homogenization techniques appears systemic in origin, not the result of a single dominant person. Therefore, a culture of uncritical tolerance likely existed for a method whose results were considered acceptable. So, why did they consider acceptable the warming bias that was somehow built into the illogical circularity of the homogenization process?
My thought is they knew of the warming bias and they possessed an ‘a priori’ premise that there must be that kind of warming from AGW. So I am led to think they embraced the process with its warming bias because it showed them what they already assumed / believed / desired to exist.
That is classic confirmation bias. Remember, this is NOAA & NASA having such a bias not a lone scientific researcher. There appears to be a disturbing problem in the climate science community’s ability to self correct systemic faults in several major US government scientific institutes.
The climate science community has been lazy. They need to revitalized themselves to take on some serious self correction of scientific faults in some major US government scientific institutes.
John
***
Victor Venema says:
October 16, 2012 at 7:50 am
The climate trend in the centre of London and Vienna is about the same as the rural trend outside the city.
***
Even if I accept that, there ain’t many truly rural sites near London or Vienna. A comparison of American cities that still have at least some such sites near them would be most instructive. Wash DC, New York, etc, wouldn’t fit that description as suburbs have spread out for scores of miles.
If a core of a large city is relatively unchanged, surrounding/spreading suburbs are still going to increase UHIE there, tho at a lesser pace. Do you think, for example, 1-2C heated air immediately upwind of a city-core isn’t going to affect it? I agree the effect would be diluted somewhat, but certainly not completely.
@beng. The region that will influence the temperature due to the UHI effect at the station is limited. Air unaffected by the UHI is continually mixing in from above. Furthermore, the higher temperatures lead to stronger gradients and thus stronger cooling by heat radiation, turbulence and convection.
I know of an empirical and a recent modelling study that showed that the heated air from Utrecht (a city in The Netherlands) can affect the temperature at De Bilt (the home of the Dutch weather service) if the wind is right. That is a distance of 3 kilometre. That may be a good estimate of a typical footprint around a station (the area that can influence the temperature). If the footprint would be less the modelling study would not have found an effect. If a typical footprint is much larger, it would have been obvious that Utrecht is a problem and that De Bilt should be treated as an urban station and these studies would not have been needed.
@John Whitman: I tend to agree. While many methods in use are superficially plausible, such as selecting proxies by correlation and reference homogenization, close examination reveals the logical flaws, and testing on simulated data shows high false alarm rates. While it is possible to do the analysis without circular reasoning, the climate community shows little interest in improving the fundamentals of their analysis and are satisfied with empirical testing of new, more ad-hoc methods, especially if the methods appear to enhance global warming. The homogenization methods in use now are a mess of ad-hockery. Some methods such as pairs analysis seem to significantly diminish the false alarm rate, but my interest would be in “why?” and whether that fits into a standard theoretical framework as recognized by statisticians.
@Victor Venema: You said: “I think the simple graphical examples on my blog on the fundamentals of relative homogenization show that this is not circular reasoning.” Your simple example does not show its not circular reasoning at all.
You responded previously that you think it perfectly justified to adjust the trend of any deviant temperature record to match the trend of some other sites. So I presume that when you go to the official Australian Bureau of Meteorology website and look up Deniliquin, for example, you would believe the strongly increasing trend and say that it is warming, even if in reality, a perfect thermometer record for the last 100 years would show a cooling trend?
Most people would have a less charitable take on that, and question the veracity of the BoM. You say the data needs to be changes according to the context of the study – biodiversity, climate change, whatever. Perhaps the data should have a label “Warning – this data only to be used for climate change studies.” \sarc off.
laterite (David Stockwell),
That your false alarm rate is high is because you did not use a reference that is representative for the regional climate around the station you are testing. Using the mean temperature series for all of Australia as a reference is simply wrong.
In a recent paper where the most used and the most advanced homogenization algorithms were blindly validated, we also computed the false alarm rate (probability of false detection (POFD), see Table 9 of the article. Except for two methods they were at or below the traditional 5% level.
I do know why pairwise algorithms have a lower false alarm rate (FAR). The FAR is about 5% for the detection of breaks in the pairs of stations. After the detection in the pairs, you still have to attribute the break to one of the stations of the pairs. This attribution is only possible if a break is found in multi pairs. For the pairwise homogenization method of NOAA, the standard condition is that two breaks have to be found on exactly the same date. As a consequence, the FAR for this algorithm was below 1%.
No one prevents statisticians from working on homogenization of climate networks. I know two of them, it would be nice if more would devote their time to this beautiful statistical problem.