From the “we told you so” department comes this paper out of China that quantifies many of the very problems with the US and global surface temperature record we have been discussing for years: the adjustments add more warming than the global warming signal itself
A paper just published in Theoretical and Applied Climatology finds that the data homogenization techniques commonly used to adjust temperature records for moving stations and the urban heat island effect [UHI] can result in a “significant” exaggeration of warming trends in the homogenized record.
The effect of homogenization is clear and quite pronounced. What they found in China is based on how NOAA treats homogenization of the surface temperature record.
According to the authors:
“Our analysis shows that “data homogenization for [temperature] stations moved from downtowns to suburbs can lead to a significant overestimate of rising trends of surface air temperature.”
Basically what they are saying here is that the heat sink effect of all the concrete and asphalt surrounding the station swamps the diurnal variation of the station, and when it is moved away, the true diurnal variation returns, and then the homogenization methodology falsely adjusts the signal in a way that increases the trend.
You can see the heat sink swamping of the diurnal signal in the worst stations, Class 5, nearest urban centers in the graphs below. Compare urban, semi-urban, and rural for Class 5 stations, the effect of the larger UHI heat sink on the Tmax and Tmin is evident.
In Zhang et al, they study what happens when a station is moved from an urban to rural environment. An analogy in the USA would be what happened to the signal of those rooftop stations in the center of the city, such as in Columbia, SC when the station was moved to a a more rural setting.
U.S. Weather Bureau Office, Columbia SC. Circa 1915 (courtesy of the NOAA photo library)Here is the current USHCN station at the University of South Carolina:The Zhang et al paper studies a move of Huairou station in Beijing from 1960 to 2008, and the resultant increases in trend that result from the adjustments from homgenization being applied, resulting in a greater trend. They find:
The mean annual Tmin and Tmax at Huairou station drop by 1.377°C and 0.271°C respectively after homogenization. The adjustments for Tmin are larger than those for Tmax, especially in winter, and the seasonal differences of the adjustments are generally more obvious for Tmin than for Tmax.
The figures 4 and 5 from the paper are telling for the effect on trend:
Huairou station and reference data for original (dotted lines) and adjusted (solid lines) data series during 1960–2008. The solid straight lines denote linear trends
Now here is the really interesting part, they propose a mechanism for the increase in trend, via the adjustments, and illustrate it.
They conclude:
The larger effects of relocations, homogenization, and urbanization on Tmin data series than on Tmax data series in a larger extent explain the “asymmetry” in daytime and nighttime SAT trends at Huairou station, and the urban effect is also a major contributor to the DTR decline as implied in the “asymmetry” changes of the annual mean Tmin and Tmax for the homogeneityadjusted data at the station.
In my draft paper of 2012 (now nearing completion with all of the feedback/criticisms we received dealt with, thank you. It is a complete rework. ), we pointed out how much adjustments, including homogenization, added to the trend of the USCHN network in the USA. This map from the draft paper pretty much says it all: the adjusted data trend is about twice as warm as the trend of stations (compliant thermometers) that have had the least impact of siting, UHI, and moves:
The Zhang et al paper is open access, an well worth reading. Let’s hope Petersen, Karl, and Menne at NCDC (whose papers are cited as references in this new paper) read it, for they are quite stubborn in insisting that their methodology solves all the ills of the dodgy surface temperature record, when it fact it creates more unrecognized problems in addition to the ones it solves.
The paper:
Effect of data homogenization on estimate of temperature trend: a case of Huairou station in Beijing Municipality Theoretical and Applied Climatology February 2014, Volume 115, Issue 3-4, pp 365-373,
Lei Zhang, Guo-Yu Ren, Yu-Yu Ren, Ai-Ying Zhang, Zi-Ying Chu, Ya-Qing Zhou
Abstract
Daily minimum temperature (Tmin) and maximum temperature (Tmax) data of Huairou station in Beijing from 1960 to 2008 are examined and adjusted for inhomogeneities by applying the data of two nearby reference stations. Urban effects on the linear trends of the original and adjusted temperature series are estimated and compared. Results show that relocations of station cause obvious discontinuities in the data series, and one of the discontinuities for Tmin are highly significant when the station was moved from downtown to suburb in 1996. The daily Tmin and Tmax data are adjusted for the inhomogeneities. The mean annual Tmin and Tmax at Huairou station drop by 1.377°C and 0.271°C respectively after homogenization. The adjustments for Tmin are larger than those for Tmax, especially in winter, and the seasonal differences of the adjustments are generally more obvious for Tmin than for Tmax. Urban effects on annual mean Tmin and Tmax trends are −0.004°C/10 year and −0.035°C/10 year respectively for the original data, but they increase to 0.388°C/10 year and 0.096°C/10 year respectively for the adjusted data. The increase is more significant for the annual mean Tmin series. Urban contributions to the overall trends of annual mean Tmin and Tmax reach 100% and 28.8% respectively for the adjusted data. Our analysis shows that data homogenization for the stations moved from downtowns to suburbs can lead to a significant overestimate of rising trends of surface air temperature, and this necessitates a careful evaluation and adjustment for urban biases before the data are applied in analyses of local and regional climate change
Download the PDF (531 KB) Open Access
h/t to The Hockey Schtick
=============================================================
UPDATE 1/30/14: Credit where it is due, Steve McIntyre found and graphed the physical response to station moves three years ago with this comment at Climate Audit.
Here’s another way to think about the effect.
Let’s suppose that you have a station originally in a smallish city which increases in population and that the station moves in two discrete steps to the suburbs. Let’s suppose that there is a real urbanization effect and that the “natural” landscape is uniform. When the station moves to a more remote suburb, there will be a downward step change. E.g. the following:
The Menne algorithm removes the downward steps, but, in terms of estimating “natural” temperature, the unsliced series would be a better index than concatenating the sliced segments.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
My biggest problem with homogenization, is that it implies that the warming from UHI can be detected as individual events. Let’s suppose that it is true.
If a group of scientists were going to measure the temperature at the center of all the baseball fields in North America, would they find a UHI bias? I guess they should not, because it is probably impossible to detect the construction of a single house at the edge of a baseball field if you only have a temperature record in the middle of the field.
In fact, you can probably build a good dozen elements of urban development either on the first row of houses at the edge of the field or inside the field. So if each of those elements added at least 0.2°C, that would amount to a good 2.5°C total. That’s as much as the UHI in the most cases. So if all the UHI is due to individually detectable events, I guess the temperature in the middle of a baseball field should be independent of the level of urban development around the field, even when there is a lot of wind.
I think I will need some empirical evidence that homogenization has any value before I can believe any conclusion based on it.
Tom Stone says:
January 30, 2014 at 8:34 am
“When a climate scientist homogenizes and smooths data, he gets a grant. When an accountant homogenizes or smoooths data, he gets gets sued and goes to jail.”
When an engineer ‘smooth’s’ the signal you get to connect to the Internet.
What is the trend for raw, unadjusted RURAL-only Class 1and 2 long-term stations within the US? Can someone make a graph of that over time or provide the data?
I think first to identify this problem was …..
Note that he is essentially agreeing with Mosh.
As for me, the short answer is that if a station is moved, I drop it.
But each of your short trends still contains UHI, no improvement here with BEST.
Yes, they will contain whatever they contain. But those issues need to be addressed (or not), segment by segment.
ScottR says:
January 30, 2014 at 8:50 am
“What is the trend for raw, unadjusted RURAL-only Class 1and 2 long-term stations within the US? ”
There is no single trend as such. Or rather an OLS trend line will tell you nothing about what s going to happen in the future.
A continuous function, such as a filter, will give a set of patterns that may, or may not, repeat exactly.
They are much more likely to give you an outline for future evolution.
So the question becomes when were these used in each weather station?
Hazen Screen, CRS (i.e., Stevenson Screen), MMTS (a/k/a “Mickey Mouse”), ASOS, AWOS, and “other” (mostly, but no exclusively Davis Pro), they all operate on the Min-Max principle. So they are all subject to TOBS bias if the Time of Observation changes.
The CRN’s PRTs are round-the-clockers, though. With triple-redundancy. But they were established after the trend stopped (starting in 2002 and growing from that point). Therefore they are too “young” to hang a trend on.
The newer USHCN systems, MMTS/Nimbus-plus-the-new-toys, and ASOS/AWOS, have automated data gathering. So you don’t have to get up at 6:00 (one of THE WORST times for observation, BTW) anymore.
But it’s all the same Min-Max principle. (And that principle has slashed my dataset and gives me no end of headaches.)
A heat sink, used in most electrical equipment, conducts heat AWAY from a component which may overheat without it. In other words, the excess heat goes into the “sink” and down the drain, away.
We call it that in our neck of the woods to distinguish it from the more generalized term of heat “source”. A driveway is a heat sink because it does not generate heat. It just soaks it up during Max time and exudes it during Min-time. A steam pipe exhaust in a WWTP is merely heat source.
UHI, specifically, is both.
In our world, a heat source exaggerates the offset more, but a heat sink exaggerates the trend more (presuming there is a trend). For this reason, Class 3 and 4 stations, on the whole, exaggerate trend more than Class 5, though the latter do exaggerate trend somewhat. A heat source that is not also a sink may even dampen the trend because the waste heat is an overwhelming factor (I think).
If a station move causes a step change, sure, that will certainly have an in-your-face impact on trend, of course. (See Mosh.)
I maintain that apart from homogenisation, questions remain about the accuracy of original observer recordings due to their propensity in the old days to round temperatures to the nearest (or lowest) .0F.
The short answer to that is “oversampling”. It still is rounded off that way even today. I’ve seen the B-91s and B44s.
Of course, “in the old days” there were fewer samples.
Instead of just acknowledging UHI effect has corrupted the historic temp data and leaving it at that,these powers (NASA, NOAA etc.) have the hubris to change these data to what they think it really should have been (after removing an unknown, unquantified UHI influence).
Once they decided to change the data. It was all fiction. Their “frames” biased these changes. How could they not do, given a “Of course the past was cooler, Global Warming is Real.” mindset.
Now, only the RSS satellite temperature data can be trusted. period.
Let’s pray the Powers don’t start changing these data.
What is the trend for raw, unadjusted RURAL-only Class 1and 2 long-term stations within the US?
+0.145C per decade Tmean (+/-0.0366, standard error), 9-area gridded average for study period 1979-2008 For Majority MMTS, with MMTS upward bump added in. (TOBS and moves accounted for.)
+0.124 Raw, but that is without accounting for MMTS conversion and you really do need to do that. Not all adjustments are out of court (unfortunately).
(That being the current result.)
ScottR says: @ur momisugly January 30, 2014 at 8:50 am…
Try this http://wattsupwiththat.com/2012/07/29/press-release-2/
Gail: Bear in mind that those results are before we addressed the (valuable, albeit smarmy) criticisms.
evanmjones says:
January 30, 2014 at 9:58 am
“+0.145C per decade Tmean (+/-0.0366, standard error), 9-area gridded average for study period 1979-2008 For Majority MMTS, with MMTS upward bump added in. (TOBS and moves accounted for.)
+0.124 Raw, but that is without accounting for MMTS conversion and you really do need to do that. Not all adjustments are out of court (unfortunately).
(That being the current result.)”
A Linear trend has no real meaning outside of the data range it is drawn from.
It cannot, and does not, provide future (or past) information.
Willis Eschenbach says: Stop boring us with your endless stories about peoples’ degrees and the like, and start talking about the science.
++++++++++++++++++++++++++++++++++++++++++
Well said, Willis.
Errrr, I thought that was the idea. If politicians wanted cooling, we would get cooling.
Richard D says:
January 30, 2014 at 10:18 am
“If politicians wanted cooling, we would get cooling.”
Politicians in a democracy rarely want to increase direct or indirect taxes. Get’s WAY too difficult to get re-elected.
That’s why Cameron (UK) is reputed to have said ‘get rid of that green…..’.
Jimbo says:
January 30, 2014 at 10:30 am
Oops that was Jimbo not RichardD -sorry.
They remove the outliers from the station population. It is the most significant cause of the great dying of the thermometers.
That applies far more to GHCN than it does to USHCN. In the latter case it is generally stations that have been closed for many years but are still part of the record that are removed. About 50 were replaced from USHCN1 to USHCN2 out of 1200+. I have surveyed, I guess, at least two thirds of the new stations. They appear, on the face of it, to be around as bad as the old ones, in terms of siting. (Long live the new boss, same as the old boss? Maybe. TBD.)
But, yes, what you say is, in essence, correct for GHCN — I think — having not surveyed them myself (yet). And when most of the stations are poorly sited and therefore reading spuriously high, there is a distinct tendency to identify as outliers and remove the good stations and promote the bad. This has a profound effect on trend, quite apart from step changes in offset. That stipulates that bad microsite affects not only offset, but trend, which is what we are hypothesizing. But you know this already. That wouldn’t matter so much if the majority of sites were good and therefore it was the bad ones being identified as outliers.
Night. Dark everywhere, except a light pole where a drunken man is searching for something. Another man is approaching and having pity with him asks:
– What have you lost?
– My keys – managed to say the drunken man.
After some time of vain search the man is asking the drunken again:
– Are you sure you lost the keys here?
– No – says the drunken man – I must have lost them somewhere there – pointing towards a dark part of the alley
– So why do you search here?
– Well, here is light, I am not crazy to search in the dark? – answers the drunken man.
I thought at this joke when realising that the paper is coming from China (who found the key in our case).
Almost all our researcher act like the drunken man researching and finding things only under the light pole of grants of research for global warming and the kind.
And doing this we waste money & time for studies that drive in circles due to our own political correct grant allocation whilst the missing heat is travelling through space for ever lost.
A Linear trend has no real meaning outside of the data range it is drawn from. It cannot, and does not, provide future (or past) information.
Yes, quite. And even linear trends themselves are, in a strict sense, statistical porn. Of which you are well aware. Yet they are not nothing.
Perhaps you misunderstand me. By “current result” I mean the result of our current study as opposed to our (lower) result in Watts et al. (2012). I don’t mean this applies to the USHCN record outside of the 1979-2008 study period, or even internal variations.
I begged and pleaded the 1979-1998 and 1998-2008 trends out of Dr. Fall back in 2010. Those results are fascinating in tandem, although the study periods are lamentably short.
[snip – too stupid and insulting to print, grow up. Plus done with an anonymous proxy and an email address that can’t be verified – Anthony]
Hi Evan,
Are the station classifications available publicly at this time?
REPLY: no, not until our paper is published, we’ve been hijacked in the past by both NOAA and BEST for sharing this data ahead of publication, and I won’t do so again – Anthony
The rural / urban distinction, as it is used by climate scientists in denying UHI influence on trends, is just silly. Effect of UHI on temporal trends does not depend on size of settlement, but on rate of urbanization, that is, changes in local log population density, down to fairly small villages.
Let’s see an absolutely rural example.
International Journal of Climatology Volume 23, Issue 15, pages 1889–1905, December 2003
DOI: 10.1002/joc.971
The urban heat island in winter at Barrow, Alaska
Kenneth M. Hinkel, Frederick E. Nelson, Anna E. Klene, Julianne H. Bell
[snip Evan – the insulting and juvenile comment has been deleted since it came from a fake email address and from a proxy server, this is a known troll trying to weasel in, so let’s not engage him, sorry – Anthony]
REPLY — So be it.