Former Virginia State Climatologist Patrick J. Michaels wrote an op-ed about his paper with Ross McKitrick from Canada’s University of Guelph in an American Spectator column today about the surface temperature record. This paragraph really caught my eye: “Weather equipment is very high-maintenance. The standard temperature shelter is painted white. If the paint wears or discolors, the shelter absorbs more of the sun’s heat and the thermometer inside will read artificially high. But keeping temperature stations well painted probably isn’t the highest priority in a poor country.”
The Stevenson Screen experiment that I had setup this summer is living proof of this.
Compare the photo of the whitewash paint screen on 7/13/07 when it was new with one taken today on 12/27/07. No wonder the NWS dumped whitewash as the spec in the 70’s in favor of latex paint. Notice that the Latex painted shelter still looks good today while the Whitewashed shelter is already deteriorating.
Click image for larger view
Click image for larger view
Whitewashed Screen on 7/13/07
Whitewashed Screen on 12/27/07
The whitewash coating I used was from a formula and method provided to me by a chemist at the US Lime Corporation, who is an expert on whitewash. He said the formula was true to historical records of the time when whitewash was used on the shelters. I was amazed to find that after just a few short months, my whitewash coating had lost about 40-50% of it’s surface area. Perhaps there was a mistake in the formula, or perhaps whitewash really is this bad at withstanding weathering.
In any event the statement of Patrick Michaels “Weather equipment is very high-maintenance. The standard temperature shelter is painted white. If the paint wears or discolors, the shelter absorbs more of the sun’s heat and the thermometer inside will read artificially high.” seems like a realistic statement in light of the photos above. The magnitude of the effect in the surface temperature record has yet to be determined, but it seems clear that shelter maintenance, or lack thereof, is a significant micro-site bias factor that has not been adequately investigated nor accounted for in the historical temperature record.
I’ll have more on this experiment soon including temperature time series graphs showing the difference between bare wood, latex painted, and whitewashed shelters.
Climate Audit 
47 Comments
Whitewash was more usually designed for mud and brick coating than for wood. It penetrated the near-surface of earthy materials better and hung on. It is well known historically to need frequent maintenance. I suspect that apart from artists, there were few users of while paint when Stevenson screens etc were invented. Lead oxide pigments were used for white from 200 AD. Zinc oxide in 1870s. Titanium dioxide (often mixed with zinc oxide for strength) about 1920s.
CA had a reference to a June 1998 paper by Jane Warne, Physics Laboratory, BOM, Australia – A PRELIMINARY INVESTIGATION OF TEMPERATURE SCREEN DESIGN AND THEIR IMPACTS ON TEMPERATURE MEASUREMENTS.
I have tried to get this but met a brick wall (unpainted).
Sorry – correction – the paper is on CA but I could not get raw data by the methods I tried, like writing to the author. Geoff.
I read a design note for heatsinks in an RCA power semiconductor data book many years ago. From what I remember the texture of the surface was more important than the colour as far as radiating heat was concerned. I think oil paint gave the best results and it specifically stated ‘any color’. This is due to the increased surface area. Although a paint is white in the visible spectrum it may not be ‘white’ at the IR end. This may have a significant effect on the rate of cooling when the sun goes down.
#1 — Geoff, you can get that report here (pdf download), or see the html version here
If those don’t work, I have the pdf and can send it to Steve M. for you.
#2 — Oops, sorry Geoff, I didn’t read #2 before posting #4.
Anthony, is it possible that the whitewash might have lasted better if the underlying wood had been slightly roughed with light sandpaper before application? Could Stevenson Screens have previously been roughened before whitewashing?
No dude that is exactly right. Every fence I ever whitewashed looked like that.
We loved whitewash. slop it on. come back in three months and slop more on.
Questions huck?
From my childhood, some fifty years ago, I can remember sitting in the outside toilet in the middle of winter warmed only by the heat of a paraffin lamp and passing the time by picking the distemper off the walls. Once the damp set in this had little or no adhesion to brick or wood. Distemper and/or whitewash were the norm for outbuildings and had been for decades. I don’t know if anyone made a distinction between the two.
Could not a simple experiment demonstrate all of this.
1) Find a particular location you think is dodgy.
2) Find another location near it free of contamination (a park perhaps)
3) Setup a simple instrument in the second location.
4) Wait 6 months and compare the values.
Maybe its been done before (perhaps there are examples of two stations close to each other but in different level of urban environments). Anyone know?
Do the site records give any indication of when the enclosures were painted?
If so, there may be a way to discern a small jump in the records at those times.
#10
The smell of the parafin and the flickering light and the howling wind. Might even build one at the bottom of the garden for a bit of escapement.
Many of the screens were replaced c. 1985 by paintless MMTS with white plastic housings. Theoretically these avoid maintenance issues, but in the thread on Sensor Blackening, at http://www.climateaudit.org/?p=2222, I argue that soil accumulation may nevertheless darken them and make them run hotter during the day. I’m not sure what the effect on nighttime IR radiation and therefore cooling would be.
Clayton B’s graph at #28 of that thread shows that CRN 1-3 MMTS sites have actually cooled relative to all CRN 1-3 sites by about 0.08 dC since 1985. (Ignore the portion of his plot before 1985, since there were no MMTS then.) If about half of sites are MMTS, that would mean that 1-3 MMTS sites have cooled relative to CRS screen site by about 0.16 dC since 1985. (I.e., screens have warmed by 0.16 dC relative to MMTS’s.)
Mark W, #9, writes,
If not, there may still be a detectable small jump in the record that would be interpreted by modern adjustment programs as an “undocumented discontinuity”, and removed adjusted out of the record. The record would therefore retain the gradual heating from deteriorating whitewash, but not the offsetting occasional repaintings. Instant GW!
Re 8, 13
I have not yet found any evidence of a NOAA database that logs application of paint or whitewash to Stevenson Screens. Maintenance was mostly ad hoc, and a lot of it left up to the individual COOP volunteer that hosted the station.
Any bias from fading or dirty paint/flaked paint/flaked whitewash is likely gradual and hard to detect. When I have my data analysed from the NIST datalogger I ran for 4 months I’ll know more.
RE: 3
Wansbeck,
With Stevenson Screens the issue is heat absorption, not heat radiation. The color certainly has a major impact on the absorption of heat energy. Modern white paints (TiO2 pigment) is reflective in the visible and IR range.
#14 Brooks,
I lived with a beekeeper for about a year and his favored paint for hives was TiO2. The reason? It is a good reflector of UV which the bees use as a homing signal.
re #14 Brooks
Surely the final temperature depends on both. Energy is absorbed, mostly, in the visible spectrum and radiated as infra-red. Also, the night time cooling is by radiation and convection. It may be a small difference but 10% can give a 0.1C difference after 5 time constants for a 25C temperature drop.
I thought it had been agreed that whitewash is more absorbent to infra-red than oxides. Have I missed something here?
the paint changed in 1979. this creates a warmer enviroment for the Lig thermometer in the
stevenson screen. In 1985 the switch to MMTS begins. so one can compare MMTS to the remaining
LIG from 1985 on. The difference between LIG ( and latex paint) and MMTS ( plastic shelter)
is shown above
There are a couple of issues associated with the surface temperature record that I have not yet seen discussed. Both of these are based on the assumption that the recorded surface temperature numbers are correct.
The first issue is the question, Is the surface temperature record the proper representation of the theoretical quantity associated with the radiative equilibrium approach to estimations of The Global Average Temperature, (The Grandest Totem with the Greatest Mojo Ever). That is, does the radiative-equilibrium approach involve the surface temperature at the altitude of the surface station measurements, or a temperature located upward in the atmosphere somewhere other than the surface? A somewhat related issue is the representation of the topographical features of the surface of the Earth used in the GCMs. Are these sufficiently refined so that the altitude of Denver, say, is higher than the altitude of the plains East of Denver. The altitude being, in my opinion, a first-order variable in the surface temperature record.
The second is associated with the comparisons of GCM calculations with the surface temperature record. No GCM has an energy equation, or Equation of State, discrete grid sufficiently refined so as to have a node anywhere near the surface of the Earth. How are the calculated numbers in the node nearest the surface extrapolated down to the surface. There are any number of ways to carry out this extrapolation. I think that maybe this issue is closely related to micro-site characteristics relative to the assumed algebraic procedures applied to the temperature distribution in the GCM calculations. That is, the assumed velocity and temperature distributions are very likely based on empirical data for a limited range of flow field characteristics and these are very likely not those encountered for many surface station locations.
I hope I have stated these clearly and they are not off thread.
All corrections are appreciated.
anthony says:
Ill have more on this experiment soon including temperature time series graphs showing the difference between bare wood, latex painted, and whitewashed shelters.
jax says:
it certainly looks like there should be a difference between the shelter readings. can you present what you have so far? even after a few months there should be a discernible difference to make it worthwhile to pursue further.
thanks.
After a couple of months of following the discussion on measuring temperatures, I am having doubts that estimating average temperature is a worthwhile indicator (and probably not an achievable indicator) of what is happening to global climate. I wonder if some other measure like the growing season lenghth (between killing frosts) would be a worthy indicator. Of course, the UHI affects the growing season inside urban areas — when September comes around, the TV weatherman is apt to warn us that “suburban outlaying areas” need to cover their flowers while gardens inside the city have no concern at the moment!
RE14 Brooks and 16, Wansbeck
See this: http://gallery.surfacestations.org/UCAR-slides/page6.html
Which is a graph of WOOD SLAT temperature over 24 hours taken with an NIST calibrated thermistor inserted in a borehole in the center of the slat.
This is not inside the screen but inside the wood.
And this change in paint caused Lonnie Thompson’s glaciers to melt, through a suburban-alpine teleconnection?
Just keeping things honest 🙂
re #14 Anthony,
The surprising thing here, at least to me, is the temperature difference at night. It seems too big to be accounted for by height and too constant for evaporation. Does anyone have an explanation?
For those who followed the earlier discussion regarding my paper with Pat M, the only substantive criticism (in my view) was the possibility that spatial autocorrelation affects our inferences. I have written a short paper on this, posted at http://www.uoguelph.ca/~rmckitri/research/jgr07/jgr07.html (scroll down the page). The test statistic for spatial autocorrelation is insignificant, and if a spatial AC term is added to the model anyway, the conclusions are upheld.
I submitted the paper to JGR, but they said that it is, in effect, a response to a comment. Since no one has submitted a comment on our paper, they cannot process the response. Despite everyone over at RC being absolutely sure that our paper is flawed, and that spatial dependence is the specific, fatal flaw in our paper, no one has written up the argument as a technical comment and submitted it for peer review. I am wondering if they really believe their own argument. John V., would you be willing?
There were some other criticisms, such as the possiblity of spurious correlation, etc, but these were already dealt with in the paper, and the feedback on these topics did not take into account the various specification tests in the paper.
RE 24. Ross did you see Gavin’s wager that if you tested your approach with GCM output
your claims would be exposed as a statistical fluke?
I started a defense, but the holidays ate my homework, you might have a look back
at the thread over there.
Folland 2001 used GCM outputs to verify the bucket corrections. Probably the biggest adjustment ever, change-in-energy-wise.
Steven, your comment presumes the usage of LIG thermometers prior to the MMTS. In fact, thermographs and other thermal recorders were in use in all of the networks, and some organizations preferred their use instead of LIG thermometers. See the station histories. Even the COOP networks permitted and used the thermographic recorders when requested and approved. Thermographs and similar devices had their own maintenance problems and error conditions.
Steven, I saw it, but de Laat and Maurellis already tested that and showed there is no predicted relationship between the industrialization pattern and the temperature trend field. In other words, the contamination pattern observed in the data is not predicted by GCMs. If Gavin didn’t believe it from them, he wouldn’t believe it from me. Also, in our paper we point out that if such a fluke correlation pattern did exist in GCM output and was not controlled for in signal detection studies, then ascribing temperature changes to GHG+sulfate is simply begging the question, i.e. assuming the conclusion they are trying to prove.
To wit: GHG forcing drives climate change. How do we know? Because there’s no way to explain the variation in the climate data except using GHG forcing. How do we know the data are clean? Because they aren’t contaminated with non-climatic changes. How do we know? Because the variation patterns are climatic. How do we know? Because GHG forcing drives climate change. [repeat as necessary]
re 27. You miss the point entirely.
if your examine all the station histories ( it takes a while ) you will see the
changes in equipment. If you want to claim that noaa gets this wrong, fine.
Yes, I’m well aware of the use of other instruments within the CRS. that is beside the point on your head.
the sites in my analsysis are divided into two groups.
1. Used a Lig instrument in a stevenson shelter AFTER 1985 , per the history files provided by noaa and the
audit conducted by Anthony.
2. Transitioned to a MMTS AFTER 1985, per the history files and per the audit conducted by Anthony
So, see that corner? see this pointy hat? Go sit over there and wear it.
It is crystal clear that Gavin takes his potshots, and mendaciously manipulates/deletes relevant posts from the safety of his RC blog, and that he will never voluntarily step up and debate in a neutral, moderated venue.
Why not?
Because RC has an agenda: AGW.
[snip]
Having worked for 30 years in a major calibration lab with all testing traceable to N.I.S.T. [and formerly to the National Bureau of Standards], I know for a fact that the ‘data’ from the surface stations pictured would be laughed out of the lab for failure to correct the underlying metrology.
[snip]
I read the graph and understood you were making a comaparison between the LIG thermometers and the MMTS instruments using 1985 as the starting point for your graph. So, I did get your point, understood your point, and raised you another point.
I was reminding the readers that the use and influence of thermographs and like instruments for measuring temperatures in the earlier decades of the weather records seems to be neglected in these discussions about the accuracy of the temperature measurements. Although you are aware of their usage, your remarks did not remind readers how a substantial portion of the past seventy plus years of temperature records were obtained by the usage of these instruments which go virtually unmentioned in these discussions. My remarks were not critical of your analysis. They were supplementing your analysis and pointing to the need for an accounting of accuracy for these instruments and their shelters in addition to such inquiries about LIG thermometers.
Inaccuracies in the station histories and the instruments used are no less prevalent than the inaccuracies about the station movements and locations. Such errors exist in the records, but whether or not they are serious enough to have any impact on nominal climatology is of course speculative and open to conjecture. A question of whether or not any such inaccuracies have anything to do with your analysis and chart was not the subject of my remarks. Again, I was only supplementing your comments to remind readers that there is a very large body of earlier data which was not acquired by the instrumentation and instrumentation problems you were talking about. As important as the question is about the performaces of MMTS instruments and LIF thermometers in CRS, there is also a very substantial body of temperature records and neglected need to evaluate these other instruments and their problems with accuracy of mesurements.
No, that’s alright, Steven. It’s your point, and I’m handing it and the hat back to you with best wishes. You’ll need it come New Year’s Eve, along with some party favors (smile). My comments about your remark weren’t about your experiment at all.
Anthony mentioned how he used a thermistor in a hole bored into a wood slat for his measurements. Your comments about the LIG thermometers and MMTS along with Anthony’s thermistor prompted my reminder about the neglect of these other instruments in this and many other threads.
Finally the answer reveals itself, and it turns out Hansen was right all along. 1998 was the hottest year.
What happened in the 1930’s? Remember the Great Depression? Obviously we could not afford whitewash for the shelters.
So since the wood became exposed temperatures went up and they went into the historical record inflated.
Now all NASA-GISS and Hansen has to do is correctly adjust for the WHI(Whitewashed Heat Island) and bingo-
All 1930 data can be shown to be exaggerated and incorrect. Besides you can’t expect poor people to get correct temp readings can you?
RE 31. At least you have a sense of humor. a Good one. Happy new year!
Anthony:
I know this prolly doesn’t really matter much, but the Stevenson screens seem to have been moved farther apart in the December picture. Why the move? Did you think the middle shelter was shielded from wind by the other two in your original set up?
Just curious.
Haven’t seen this paper mentioned yet:
http://cdiac.ornl.gov/ftp/ndp070/ndp070.txt
UNITED STATES HISTORICAL CLIMATOLOGY NETWORK
DAILY TEMPERATURE, PRECIPITATION, AND SNOW DATA
C. N. Williams, Jr., R. S. Vose,
D. R. Easterling, and M. J. Menne
National Oceanic and Atmospheric Administration
RE34 Dan, good eye.
The picture from 7/13 does show shelters closer together, and I realized that I had made a shading issue unintentionally. I hadn’t started data logging yet.
I moved the shelters apart a couple of days later. I started my first run on 7/20 and then had a data link failure. All the instrumentation problems were solved in the next week and the first full day of datalogging was on 7/28 in the position they are in now and seen in the 12/27 picture.
I ended the datalogging earlier this month and I’m now starting on the data analysis.
Re # 18 Dan Hughes
Can’t agree with you more. I’ve posted similar material over the months, but so far as I can work out the “official” climatology logic seems to go –
we need a model to look into the future
models are bounded
therefore, we need data from past climates
but, the data are imperfect
so, let’s adjust them as much as we can
to discover, that the errors are greater than the data
but that, we need the data whatever the errors
otherwise, we are out of a job.
Errors include both conceptual (how high above the ground is representative for temperature, if any height is?) errors of physics (Le Chatelier indicated that heat tries to even itself out by hot moving to cold, but there is no equivalent Principle for say rainfall & clouds) and errors of execution (of which Steve has picked up many in mathematics).
But the biggest error for my way of thought is emerging as the error of omission, keeping inconvenient data hidden when its inclusion would alter the dogma.
As a minor aside, I floated the idea that Hg in glass thermometers in a Stevenson screen see some radiant light, as do MMTS type devices. They do not rely 100% on convection or conduction. Do they have different spectral responses to radiation, like some being sensitive in the IR and others being blind? Or is the effect too small to worry about?
Ah, but that omission is an error of commission.
=============================
#8 – that experiment might get you a difference between stations, but it won’t tell you much at all about trend.
#32 – You may have been facetious, but that is a point that I rarely see mentioned on this blog. Somehow we are concerned about a station that has been close to a road for 30 years, but are completely ignoring what is likely to be a bias from inferior monitoring, equipment and sheltering from older readings. Does anyone think that the rural townspeople were out washing down their enclosures every day during the Dust Bowl?
Visualize it: husband leaves for work, wife removes a Stevenson screen from the CRS, props the slatted Stevenson screen into a galvanzied wash tub, scrubs the laundry on the Stevenson screen, puts the Stevenson screen back into the CRS, welcomes husband home, husband makes weather observation using instruments in the CRS….
Steve
Here is an interesting paper that was presented as a print only paper at the Lunar & Planetary Science meeting this year.
ftp://ftp.lpi.usra.edu/pub/outgoing/lpsc2007/full818.pdf
Grist for the mill in looking at Urban heat island effects.
Anthony 36:
Thanks for the reply. I’m looking forward to your results!
Re: 37
Geoff,
My guess is that there is a difference and it is significant. The lack of a study comparing the two methods is another metrology failure.
Anthony,
I looked into the heat transmission through wood a while back. All the studies I found showed it was a very poor, slow conductor. Also Stevenson screens allow air to flow freely from sides and bottom so the heat that transfers through the wood must heat the air in the enclosure that can constantly mix with outside air. The effect of heat transferred through the wood and through the air to the centrally placed thermometer should be minimal with or without painted surfaces.
I think a closed box experiment would be required to see what effect applied heat to the wood enclosure had on the temperature reading inside the box. This could then be compared with the open box, free air flow readings that would include windy days, humidity, etc…
However, saturated woods conduct heat comparatively well. So deteriorated woods are suspect for more rapid and exaggerated heat transfer into the protected area of the box. I noticed in some of the photos on your site (some time ago) that some of the Stevenson Screens did show signs of rot and deterioration. This could allow the woods to saturate and lose their insulating capacity.
Why do we put up weather stations?
Benjamin Franklin had one, which is how he put weather predictions into his almanac. Generally speaking, weve probably always had weather stations, they just werent called that. Thered be one old person who had lived in a particular area for decades. They would walk outside in the morning, and there would be a certain feel to the air, and the old person would announce that it was time for planting, or the winter would go a certain way, or high winds were coming. They were a weather station that was self-calibrated to assess incoming data and deliver a report of projected weather.
Then thermometers were invented, and data was recorded. However, it was for the same purpose, which was to tell if it was time for planting or harvesting. The temperature was only important in terms of what it meant for accomplishing daily or seasonal activities. A weather station is of great practical importance for farmers.
Then airplanes came along. Local weather being also of great importance to pilots, weather stations were often put up at airstrips, which were all dirt and grass to begin with. So weather stations for farmers and pilots looked at the same background. Once airstrips started being paved and buildings being put up, it did change the nature of the weather stations, but not the essential need for them. The localized weather conditions of the airstrips/airports were still required by pilots, especially as they would be affected by those ground conditions while both taking off and landing.
I would argue, then, that it probably does not matter what conditions the weather stations exist in, as long as those conditions are what the local population wants to know about. It would be good to have some standards, just so the local weather can be reasonably compared from one season to another.
But the question still remains: Why do we put up weather stations? And the answer still remains: So that local weather patterns can be recorded, and the local people can be apprized of what can be expected, based on previous experience.
What are weather stations not? They are not good approximations of temperature, humidity or rainfall for other than local areas, their readings cannot be extrapolated out over large areas. They are not necessarily precise or accurate, especially for other than the local areas, but for what they are supposed to do they are precise and accurate enough. They are not of any use in coming up with some global average temperature (if there is even in meaning in that phrase), because there are not enough of them, they are not scientifically placed, and they are not maintained by experts who consistently follow specific instructions.
Weather stations are an excellent tool, but only for what they are best at, which is local weather data. Using a wrench instead of a hammer to drive in a nail is just bad practice. Using weather stations (in their current form) to predict global climate is simply using an excellent tool for the wrong job.
Anthony:
Let’s keep in mind that the keepers of the SATs have already conceded that the actual raw temperatures recorded by these devices are meaningless if taken as individual data points http://data.giss.nasa.gov/gistemp/abs_temp.html. If any given device is off by a degree or two on any given day, so what? The data will be “adjusted” anyway. So there are two issues: 1)reliability of the data (insignificant because the data will be smoothed and extrapolated for use in models) and 2) methods of data smoothing, UHI application etc. The second issue is the main reason why this blog exists, IMO. When a governing body (NASA) is less than forthcoming with their statistical methods, quality of data sets and scientific methodology we have a major problem. Don’t get lost in the weeds of whitewash vs. latex, the powers that be don’t really care. They’ve moved beyond such trivialities of the integrity of the data gathering process and have moved on to driving international policy.
Anthony do you know these?
Thermometer Screen Intercomparison in De Bilt (the
Netherlands)
Part I: Understanding the weather-dependent temperature differences
Short title: Thermometer screen intercomparison Part I
J.P. van der Meulen, T. Brandsma
Royal Netherlands Meteorological Institute (KNMI), De Bilt, the Netherlands
International Journal of Climatology (accepted February 2007)
Corresponding author: Dr Theo Brandsma
KNMI
PO Box 201
3730 AE De Bilt
The Netherlands
Telephone: +31.30.220 66 93
Telefax: +31.30.221 04 07
E-mail: theo.brandsma@knmi.nl
Click to access meulenbrandsma2007b.pdf
Thermometer Screen Intercomparison in De Bilt (the Netherlands)
Part II: Description and modeling of mean temperature differences and
extremes
Short title: Thermometer screen intercomparison Part II
T. Brandsma, J.P. van der Meulen
Royal Netherlands Meteorological Institute (KNMI), De Bilt, the Netherlands
International Journal of Climatology (accepted February 2007)
Click to access brandsmameulen2007a.pdf