Anthony Watts reports on whitewash versus latex.
Since they mentioned paint, I thought I’d share this experiment with everyone. You may recall that I’m conducting an experiment measuring the differences in temperature imparted by different coatins on the Stevenson Screens. Early screens were spec’d to have whitewash. I made up a batch of whitewash based on a known historical formula and setup a test rig:
Here is one day’s worth of plotted data related to whitewash -vs- latex -vs-
bare wood I have three weeks of this logged data and they all look pretty
much the same. Russ Steele was kind enough to plot the data for me for a day, May 21st.
These are temps recorded in a 1/4″ borehole in the center of the wood slat
along with air temperature from an aspirated stacked plate IR shield. The
datalogger I used is NIST calibrated, and the probes are NIST calibrated.
Error is 0.1 degree F or less, and they are quite linear.
I have made it available on
http://www.surfacestations.org/downloads/paint_test1.zip These are all comma
demilited text files, suitable for import into R or Excel. This is three weeks worth of data.
The columnar format is:
date, time, Air Temperature, Bare Wood Temp(control), Latex Paint Temp, Whitewash Temp all in degrees F
I have my three new Stevenson Screens in, so I’m going to switch now from
wood borehole tests of the painted slats to actual air temperature tests in
three standard Stevenson screens. I expect the temperature differences to be
smaller, but still measureable.
I have aquired the exact formula for lime based whitewash and the exact
materials from a chemist at the National Lime Company. They are watching
closely too becuase whitewash appears to be a superior coating for buildings
to create energy savings for air conditioning.
Will advise when the new experiment is up and running, it will be live online.
Climate Audit 
96 Comments
Ralph Becket said this in the The Minnesota Dam Nation thread:
Running through Anthony’s paint_test1 data I get
mean daily max latex_temp – whitewash_temp
= 4.1 F
mean daily min latex_temp – whitewash_temp
= -0.8 F
mean daily latex_temp – whitewash_temp
= 1.2 F
mean min latex_temp
= 53.5 F
mean min whitewash_temp
= 53.9 F
mean max latex_temp
= 96.8 F
mean max whitewash_temp
= 93.5 F
The spread for mean min/max temps for latex paint is 43.3 F; the spread for mean min/max temps for whitewash is 39.6 F.
Thanks Ralph, I welcome any independent analysis of the data that I posted
The primary results are promising.
If I was reviewing this for a journal publication, I’d have a couple of nitpicks/questions. They aren’t essential, but it always nice to be able to say “Yes, I checked that, it was irrelevant.” For one, in this sort of measurement I would mark the physical probes and rotate them. (Recording which was which, when, naturally.) With their wires. Also, there’s the physical orientations. The picture above is ordered ‘post, something-white, bare, something-white.’ Just record the order and rearrange. Randomly. Or methodically to all permutations.
I don’t _expect_ any issues, but a kink in a thrermocouple wire, or an intermittent shadow or something is easy to discern… _after_ the fact.
The key to defending things is the 7 million non-factors that you can eliminate from the core argument. (Which is exactly why the whole AGW debate is so infuriating. Every single one of the side issues was examined by someone with an axe to grind. Once. With non-public data.)
And FYI, The NWS made semi gloss latex paint official in their published: “Specification for Shelter, Cotton Region Type” SP004 in 1979.
I’m still looking for SP003 and earlier versions of this publication, if anybody has them. Anybody who finds the earlier versions gets a free CSI t-shirt.
Here is a BAMS article “Some Perspectives on Recent In Situ Air Temperature Observations: Modeling the Microclimate inside the Radiation Shields* They do comparisons on different sheilds and sensors, including MMTS.
No mention of the paint issue. Its too obscure. If you haven’t worked with the older screens, you wouldn’t even know about it, its a met-tech level issue.
http://ams.allenpress.com/perlserv/?request=get-document&doi=10.1175%2F1520-0426(2001)018%3C1470%3ASPORIS%3E2.0.CO%3B2
RE2 Al, excellent points. The CRS’s I have will not be permanently anchored, so it will be easy to rotate them and the probes.
Didn’t see this thread before I posted my graphic in the other thread. You might want to take a look.
http://www.climateaudit.org/?p=1702#comment-114644
SteveM asked me to post the data unzipped as individual text files for those that may want to use it, here it is
Just replace the number in ( ) to get different 24 hr periods.
http://www.surfacestations.org/downloads/paint_test1/TempElertRaw5-20-07(1).txt
5/19/07
http://www.surfacestations.org/downloads/paint_test1/TempElertRaw5-20-07(2).txt
5/20/07
…
http://www.surfacestations.org/downloads/paint_test1/TempElertRaw5-20-07(23).txt
6/10/07
Data is 1 minute sampled, format is:
date, time, Air Temperature, Bare Wood Temp(control), Latex Paint Temp, Whitewash Temp all in degrees F
comma delimited
It’s not clear what you are doing, and what you are hoping to find? What are you hoping to find?
From the picture and description I think you are:
Measuring ‘air temperature’ in the screen?
Then measuring the internal temperature of the bits of wood with various or no paints?
Surely, if you wish to *just test paint*, you have to compare like screens with just differing paints? You also have to expose all three screens in the same way? The blocks of wood are not, obviously, aspirated like the screens or exposed in the same place. For example, might their lower night time temperatures simply be because they are nearer the ground? We don’t know, you’ve not eliminated this question from you experiment.
That said I don’t find the responses you’ve found surprising – they’re also certain to be known already, people have done these tests in the past.
I also see the wood is nearer black stands than the screen…I may be wrong but haven’t you complained about black asphalt around some weather stations?
But, doesn’t this highlight a point? Namely that what matters is if a sites characteristics changes over time? The reading you have are comparable if (*IF*) exposure stays the same over time.
Hi Anthony, looking at the cooling of the peices I am strick that the rate of cooling of your pieces is greater than that of the air. The slope of the curves of the wood and the air are at 36 and 43 degrees. This means that we have a water phase transition involved somewhere.
Allow me to make a small suggestion. Stay up until midnight two nights running. Place wooden samples painted in different ways in an oven and heat slowly to 100 degree F (make sure they are all at the same temp). At midnight, where the night air is cool, take them out and record the cooling profile.
The following evening repeat the procedure, but at the base of the oven have a tray of water, so that the oven is humid. See if the paints radiate heat differently whan exposed to water. My guess is that the surfaces arre all “sweatly”.
So this topic is not about LaTeX.
Anthony,
In addition to the variations suggested by Al in his comment, I would also recommend continued testing to simulate the maintenance of the screens. After collecting data on your three sample slats, add an additional layer of whitewash and latex to your existing test samples. At the same time, create a fourth test sample that has a base coat of whitewash with a top coat of latex.
Other things I would look for are gradual changes over time like greying of the bare wood and latex paint pealing revealing bare wood underneath. You may need a few more boards than what you have though. — John M Reynolds
Anthony, nanother test (And I’m sure you’ve alreadyt thought of this) – what is the impact of placement not just near asphalt, but on a concrete slab a la Headworks Portland?
RE7, Peter. This test is just a precursor to the real test with 3 Stevenson Screens.
What was I hoping to find with this test?
1- That there is a difference, in situ, between different paints used to coat wood used for Stevenson Screens
2- That there was a significant enough temperature difference between whitewash and latex paint to measure
3- That the temperature difference was reapeatable, and it was not a fluke of drying or curing
4- Given the sum of 1-3, that it justified doing the full scale screen test, which has an out of pocket expense for me of about $3000.00. I have no grants, or backers, or government funding. I wanted to make sure before I spent this money that there was a likelihood of a measureable result. The best way to do it was a small scale test which albeit imperfect did show results.
This is not a test for a scientific publication, but a trial test to help me decide what to do next.
I also did this test a different way, using an IR thermometer (before installing the NIST datalogger) and found similar results. The IR thermometer had an emissivity setting, and was reasonably well calibrated. It showed nearly identical results for the different wood surfaces. See http://www.surfacestations.org/spot_test/paint_test.htm
In the case of the Stevenson Screen test, the conditions will be more tightly controlled, with identical screens. The air temperature is there more for reference than anything. Air temperature isn’t the driver, visble and IR radiation is.
I will have a pyranometer running in parallel for the screen test, along with air temp and wind.
“…people have done these tests in the past”
I’d sure like to see them. I’ve done a search for a couple of months on a variety of locations, including the AMS extensive online database of publications. I have not found any instance of a study done on whitewash -vs- latex paint on Stevenson screens. I did find a number of siting tests, and comparison tests between shelter types, but nothing on paint.
If anybody can show me that this experiment has already been done, please point me to it and I’ll not waste any more time on it.
Well, Peter has claimed that people have done such tests. And we are sure that Peter is not simply making stuff up, so he must have those tests in hand. Post ’em, Pedro.
Anthony, interesting reply. You clearly are looking for a certain result. I think that’s going to bias you. Why don’t you look for effects that might *COOL* Stevenson screens over time? or do you rule them out?
It may be the case you’re the first person to think about what differing paints do to Stevenson Screens, but I have to say I wonder why meteorology settled on white paint ages ago if the questions hasn’t been asked.
Re #16
Peter Hearnden,
You may of course be right, these questions MUST have been asked before. And SURELY someone has done these tests before. So do kindly point us to the questions and answers. Otherwise please do continue to identify procedural bias in these tests so that they can be accounted for and eliminated.
Re, #15.
No, I don’t have that kind of thing to hand, it’s just my understanding, something I’m pretty sure I remember reading. Implying I’m dishonest might win you friends here, it leaves me pretty unimpressed but, sadly, not surprised.
Still, now I remember why I stopped posting here and why you see zero dissent here these days. Post anything contrary to the line here and you soon get your character attacked. Yuck, I’m off.
So even if someone has done the work, there’s no point in performing the experiment yourself for verification/replication?
Typically, when someone here bemoans the results of a study, the criticism of that someone tends to be “well do your own study and see what you get.” And here we have someone doing their own little study, yet they’re being criticized for doing so.
#18. Peter, I saw no attack on your character. You were however asked to provide a citation or support for your argument. Saying that “you’re pretty sure that you read it somewhere” doesn’t cut it here, I’m afraid. Don’t blame anyone here because you failed to support your argument.
Peter,
I suggest that you go down to your next high school science fair.
Making a proposition, and then testing that proposition is the way such science is done.
If you can find a flaw in the tests, please spell them out.
Anthony has stated over and over again, that the purpose of this test is to determine what affect the changeover from whitewash to latex is on stevenson boxes. Based on the absorbtion spectra of the two paints, it has been proposed that there will be a difference.
Once again, if you have a real complaint about the tests, please let us know. If you are just complaining because you don’t like the fact that someone is challenging the assumptions of the AGW alarmists, then don’t waste your, or our, time.
Doc Martyn,
The wood should cool faster than the air initially, since it is hotter. However, I suspect that you are correct about condensation, as the slope of the wood’s decline in T decreases when it gets more than 3-4 degress cooler than the air. That would suggest to me that there is probably condensation retarding the exchange of heat between the wood and the air.
BTW I’m sure that Anthony would be equally willing to do tests for the opposite sort of bias – be it the growth of trees, whatever. While his testing may have started with paint, I’m sure that his interests are now microclimate in general and will test a variety of cases. The more that you think about it, the more unbelievable it is that USHCN does all these adjustments without experimentally testing (And reporting) what variations and biases can be expected within microclimates. Shouldn’t there be a catalog of biases – both positive and negative?
RE 16 Peter,
It isn’t that “meteorology settled on white paint” its that the paint composition and the paints applied changed over time. Early Stevenson Screens were painted with whitewash, which has a significantly different emissivity/absorptivity that latex paint, which my research shows was the NWS specification as of 1979 for Stevenson Screens.
Based on the field studies I’ve done so far, examining actual screens in the field, it is a certainty that the majority, if not all, have been repainted or commissioned originally with latex rather than whitewash. My experiment is designed to test that observation to see what it brings forth. There’s a possibility that when the full size screen test is done, the results I’ve seen with slat bore holes may disappear as there may not be significant heat transfer to air passing through the screen. The only way to know is to run the test on 3 screens.
Again if you have a reference that shows this experiment on whitewash -vs- paints was done on Stevenson Screens before, please point me to it. It would save me a lot of work and personal funds.
And if you see a potential bias brewing, I’m all for eliminating it. What effects might you suggest that I test? Testing many different ways can expose weaknesses in the experiment. I welcome suggestions from everyone, such as SteveM’s concrete slab test. Which I have a way to do.
I haven’t done the full scale test yet, so there’s plenty of time to work on eliminating any bias or mistakes. I don’t have a peer group at a university like some do where ideas can be bounced off others and tested, so I welcome suggestions and criticism.
Re #8
Peter Hearnden,
It appears you are indeed wrong. At least regarding the wood being closer to the black supports. The wood is in the center, with latex and whitewash on the sides. So the bare wood being furthest from the black supports yet still showing the highest temperature differential minimizes the argument that the black supports are influencing the readings in a significant manner. Nonetheless, whatever the effect is could be further minimized by extending the span by placing the frame on 2×4 boards.
9:
I was wondering about this, too, and I think you are right. Wood absorbs and “evaporates” water very quickly.
RE 9, 22 The chemist at National Lime Compnay told me to expect water vapor condensation and evaporation to be part of the effect. He told me that Whitewash does a superior job at this, cooling during the day due to evaporation.
RE18, Peter, I urge you not to sign off. Criticism of what I’m doing is important. If you can remember what the papers might have been titled or the authors, I certainly want to track them down.
Another thing to add to the list of possible effects to test would be whether the condensation effect changes when the broad flat surface of the slats is nearly vertical as opposed to being horizontal. If the condensation can roll off instead of pooling on a flat surface, does the effect change in magnitude?
RE23, Steve here is a paper showing a 0.3 degree centigrade positive bias on MMTS changeover from Stevenson Screens.
Click to access i1520-0477-72-11-1718.pdf
Re #9 Aren’t the slats radiating heat into space?
#22 Condensation would warm the wood tending to keep it’s temperature towards the dew point.
The difference in temperature is influenced by radiation from the ground, condensation, air flow, thermal mass of the slats, insulating properties of the wood and apparently paint.
RE: Peter and others who are “offended” by this testing. What gets them angry is the possibility that a test might be done by a sceptic, who is not of “pure mind.” In other words, even though the scientific method, if applied correctly, will overcome any bias or expectation on the part of the experimenter, now, those who defend the current orthodox view regarding AGW insist that not only must the scientific method must be followed, (or not!) but most importantly, far more important than the science itself, any experimenter involved in this issue must be of “pure mind.” Of course, the bar will be endlessly raised, and numerous red herrings thrown. And the sad thing is that, the masses, who are, in the majority, scientifically illiterate, will be suceptible to such ongoing arbitrary shooting down and any and all challenges to the orthodoxy. What sad times we live in.
Peter…this is not an attack on your character.
Null: There is no difference in microclimate between whitewash and latex paint.
Empirical data: There is a repeatable difference in microclimate between whitewash and latex paint in a preliminary study.
therefore, the null is excluded.
granted, Mr Watts probably would not have written up anything had the null NOT been excluded. other than that, could you please show how the experiment is biased?
It looks like max temp with all these systems would be quite dependant upon the amount of sunlight. You might see far different results on cloudy days.
Peter your comments raise some interesting points and like others I urge you to continue to comment.
In general, I assume there is an “experimenter” bias in the testing of any proposition. This raises the ante on clearly describing the proposition and what was done to test it so that the experiment can be replicated by other parties, including those with the “opposite” bias of the original experimenter. It seems to me that Anthony has been very explicit about what he has done and has been very open in sharing his data. It also sounds like Anthony would be perfectly ready to work with others who want to replicate the experiment or alternate experiments.
What is psychologically and sociologically intriguing here is what appears to be a basic antipathy to the testing of a proposition by someone not in the priesthood. (I am assuming that Anthony does not currently have a PhD in Climatology, Materials Sciences, etc.)
Re #30,
Anthony,
That +0.3 C bias was of the minima, not the average. The
average bias was -0.1 C because the maxima bias was -0.4 C.
IIRC, it was a statistical study comparing data from old
stevenson screen sites with data from new MMTS sites without
examination of micro site details.
Good work Anthony.
I’ll just lay out some issues coming off the top of my head.
1. You should document your test site using CRN paperwork. Take all appropriate photos
and make sure we do a CLASS 1 site as a baseline. Then one can look at SteveM suggestions
( asphalt, concrete, shrubs, etc ) and simulate a Class 5 site. This will help to address
issues like Peter raised ( the black plastic surface in your test… Not that plastic
stores heat like Asphalt or concrete, but best to keep these annoyances to a minimum)
Anyways, use THEIR method to document your test site. If its good enough for CRN…
2. Rotating sensors. I understand the concerns others raised. My concern would be that you
might induce flaws by rotating. Is there a way you can periodicly test the sensor against a reference?
or test it after rotation? I’m not talking about a full calibration.
3. Folks have speculated about a water related effect. ( doc in #9) This givs me pause.
Funny thing about your charts. I see Lower lows and higher highs in the wood structure
versus the sheild. Questons for the bigger brains
1. Since the GISS mean is the mid point of the day high and day low, the question
comes… does a wood screen bias the high in the same way that it biases the low?
2. Wet wood measurements? also, how does each paint hold the water out/water in
3. rain and temp correlations. Lets assume Latex holds the water out better than
whitewash. When it rains on each, what kind of differential effect do you see.
Same for snow. If there is a wet wood bias, how soon does it go away, so
time series correlation effects for wet wood. Simply put, The current MMTS
shield ( I assume) doesnt retain moisture in its physical structure. Wood
may react differently in a period of rain, especially if it takes up water.
So when they tested the new MMTS did they test wet and dry?
4. Wood type,wood age,wood condition. I’ll read the MTS experiment and see what they did
5. Temp excursions: Your test is being done for one set of temp excursions. Do you plan a cold weather
series? dew point? wind? soil temp, insolation measurements?
6. Layers of whitewash? Doubtless some will argue ( and it may be a good point) that IR bandpass
properties of whitewash are retained when painted over by latex, think years and years of lime
impregnated wood. Cool thing would be to get an old Screen.. or wood from an old screen..
7. Power of your test. Statistically speaking. SteveM and bigger brains can and should help here.
8. Tip Jar. Do you have one on yur site? Might be cool for folks to help fund the experiment.
Even on an ongoing basis. Kinda like sponsering a walkathon.
More later
“They are watching closely too becuase whitewash appears to be a superior coating for buildings
to create energy savings for air conditioning.”
That qualifies for a consulting fee, Anthony. You should work out a contract with them. Any money can finance your site audit. Capitalism at its best. 🙂
Methinks having additional setups run concurrently is necessary (“duplicate reactors”). I’m not sure that the “control” is…air temp? Plain wood? One control might be ok, but at least duplicates of the painted wood, if not all three. Multiple “test rigs” would be even better.
He’s not interested in anything other than questioning motives, I don’t think.
Regardless, most anyone that spends 10 seconds thinking about putting multiple coats of paint on something is going to come up with the thought “I bet that changes the temperature somehow, probably warmer. I wonder if so, and by how much.” I would think if you’re interested in finding out how accurate things are, you know, that science thing, you’d not only want this experiment done, you’d be happpy somebody was doing it, as long as it was done fairly and in an accurate manner. I wouldn’t think anyone could seriously say that Anthony isn’t doing it that way, regardless of why. Then the data is there to prove or disprove.
I think the mind-set that’s probably here; “CO2 is everything. Since there’s more, it must be warming, regardless of how many bad stations there are.” Why everyone is so scared that will be shown (rather than glad we can get rid of them and see what’s really going on) only have a limited number of reasons, the prime ones being they’re not interested in what’s really going on and they’re not interested in the science.
I can see asking questions on the specifics of why, but not being upset about answers nor asking the general of why; that’s obvious.
Keep up the good work!
Anthony,
I like what you are doing. I think it can only help in the long run. If climatologists or meteorologists disagree with your methods and results, perhaps they will be inclined to conduct their own tests. I certainly think it will make others more mindful of their own methods for testing and data collection.
One thing I notice is the slats are exposed horizontally to the sun and ponding of moisture is allowed. The photos I view of the Stevenson screens show only vertically oriented members where ponding is not possible (or unlikely). The roof is sloped and will tend to shed water.
Also the wood grain is oriented differently in your study than on the screens. I don’t know if this would make any difference, but as I only know about structural engineering, it is something I would wonder about. Orientation of grain, exposure of end grain (where moisture is readily absorbed) and slope to prevent ponding are important issues in my world.
Switching from Wood to the MMTS structure.
Anthony I looked at http://ams.allenpress.com/archive/1520-0477/72/11/pdf/i1520-0477-72-11-1718.pdf
fascinating. When they moved to MMTS the MAX and MIN for daily’s moved exactly how
your max and mins moved, when you compare the sensor in the IR stack versus sensor in the wood.
FUNNY, they did not do a pristine A/B comparison. They made the change in the feild and then tried to
back out the impact.
What is the MMTS shelter made out of? Clearly, when they shifted to it, the climate inside
the space changed. Are you using a MMTS structure?
Other thing: they note that moving to MMTS moved stations closer to building because of CABLE LENGTH.
Finally, A listing of the sites in their study would be good.
Re #16
Could you suggest a few items that might cool the screens? Here are a few I thought of: shade from growing trees eventually covering the enclosures, teenagers using the enclosure to store their forbidden beer, glacial meltwater enterig the site, nearby suburbs or industrial sites being returned to cultivated land or forest after losing population, a burnt out light bulb within the enclosure, the weather station being moved further out of town, the weather station being moved downwind of the city so it gets more cooling rain, an enclosure being painted with another coat of whitewash/paint because the first started looking tacky, a thicker coat of paint compared to a thinner one, an exhaust fan being installed to evacuate air from the enclosure, aging of the sensors, dust buildup, a leak in the roof, birds nesting in the enclosure so it was read only in the cold months when the nest was empty.
Some of the above are jokes, of course.
So what the heck is a CRS or Cotton Region Shelter?
I’m curious so after a bit of googling I found this article and it would seem that what we know as a Stephenson Screen is known as a Cotton Region Shelter to distinguish it from a Fruit Region Shelter. The FRS is more open and it would seem designed in be used in the shade of fruit orchards.
Wikipedia says the Stephenson Screen was invented by Robert Louis Stephenson’s father.
You may be finished with this stage of the experiment, but if you are going to continue a few suggestions to reduce some of the moisture issues:
1. Choose and photograph a board of consistent grain for its length.
2. Cut all pieces from the same board.
3. Orient the boards at an angle to approximate the slats of the Stevenson box. This will shed water from all faces as there are no planar surfaces. Also the angle of the sun to the board may be a factor.
4. Seal the end grain with a waterproofing sealant. The Stevenson boxes do not, for the most part have exposed end grain. Don’t use latex or whitewash for the end grain as these will allow absorbtion into the end grain at different rates, which is not consistent with the Stevenson box.
5. Don’t use a channel to support the pieces as this tends to trap moisture. Try a nail in each end or something like that.
I don’t know that it will change any of your results, but it does more closely approximate the Stevenson boxes with respect to moisture and orientation.
Anthony,
If a full scale test costs $3,000 or so, why don’t you put a contributions applet on your web site? I suspect that there are quite a few people who would send you $5 or $10.
I’ve redone the links for the data, since there was a problem with the links I posted, I also added weather data from a weather station about 100 feet away at my office, which includes solar radiation along with other variables.
http://www.surfacestations.org/downloads/paint_test1/TempElertRaw5-20-07.txt
http://www.surfacestations.org/downloads/paint_test1/TempElertRaw5-21-07.txt
…
…
http://www.surfacestations.org/downloads/paint_test1/TempElertRaw6-11-07.txt
There was a thunderstorm event with .4 ” rain at about 2PM on June 5th, and I’ve made that data available here:
http://www.surfacestations.org/downloads/paint_test1/wxdata_6-5-07.txt
and the entire set of weather data for the period of the experiment, in 1 minute intervals
http://www.surfacestations.org/downloads/paint_test1/wxdata_52007-61107.txt
The columnar format are in the weather data files at top. Ignore any “inside” values as these are not part of the experiment. The format for the Temperature wood slat probe data is:
date, time, Air Temperature, Bare Wood Temp(control), Latex Paint Temp, Whitewash Temp all in degrees F
Data is 1 minute sampled, comma delimited
Thanks for all the great suggestions.
Re46, I thought about it, but I don’t want to be criticized for financial reasons. Easier to just do the work and worrk about finances later.
Nighttime air temperature measurement – why do they remain higher than all other temperature measurements? In finer timely resolution, I can see that wood lags air by app 5 minutes and wiggle amplitude is about the same for all measurements, so they all seem to monitor the same ‘regime’.
At present, I have no other explanation than calibration offset.
RE44, backtracking from that link, I found this little page. Climatic data that has been “rescued” from around the world.
http://docs.lib.noaa.gov/rescue/data_rescue_home.html
re: 50
you’re amazing, Anthony,
for some of these stuff I was looking for quite a while.
Thanks
RE34, jae, see the parallel sampled wxdata I posted that has solar radiation in watts/meter^2
Yes the driver for this is solar energy impacting the wood. Stevenson Screens were designed to minimize direct solar, reflected solar, and radiated IR
RE42 thanks Steven, I found the NWS page that specs out the MMTS
http://www.srh.noaa.gov/ohx/dad/coop/mmts.html
And see this, some “lessons learned” about MMTS, including bees, connectors, and dog chew toys
Anthony: Good work. It is science. Try to keep it simple. Consider all ideas though you cannot try everything. Trying to incorporate every suggestion will merely ensure that nothing much gets done.
I was baffled by Peter. And quite disgusted when he suggested you should be working on something else.
i.e. #16 ‘ Why don’t you look for effects that might *COOL* Stevenson screens over time?’
That is not only a classic diversion, it is a classic of arrogance. You project is yours. And an answer is obvious; your test isn’t for effects over time, you are looking for effects from whitewash and latex.
If anyone reads this far. Some recent papers say soot and particulates have reduced the albedo of artic ice which may explain some melt. The same soot would fall on measuring stations but I have no idea if it can be having any effect.
Whoops for the link for Lessosns Learned on MMTS
http://www.srh.noaa.gov/ohx/dad/coop/cooplessons.html
RE: #54 – That may partially explain why NH has seemingly lost sea ice extent over the years. Carbon particles affect the NH more than the SH. Of course, having an ocean at the North Polar region is another key difference, there are different oscillatory dynamics affecting the Actic Ocean than those which affect the Southern Ocean.
Anthony,
I read those lessons learned the other night and had a good laugh.
Not only Latex matters:
I did go to DWD – the German Weather Service –
http://www.dwd.de/de/FundE/Klima/KLIS/daten/online/nat/index_mittelwerte.htm
to get the “internet_temp_1961-90.zip” to get the median/average temperature
between 1961 and 1990 (they don’t tell which way, but it should be the median))
I sorted the (zipped) .xls by elevation, (all of the 675 stations included, have elevation mentioned), I added some columns into that .xls, created
quarterly averages from the monthlies. Then I divided stations into classes
of eleveation (0 – 69, 70 – 139, ….., elevation in meters)
Elevation(m) Stations(sum) Year DJF MAM JJA SON
0-69 146 8,6 3,3 7,8 16,3 9,4
70-139 79 9,2 4,0 8,7 17,3 9,7
140-209 61 9,1 3,6 8,8 17,1 9,4
210-279 57 8,8 3,2 8,4 16,9 9,2
280-349 57 8,1 2,5 7,8 16,4 8,4
350-419 59 8,0 2,3 7,8 16,4 8,4
420-489 57 7,8 2,1 7,5 16,1 8,3
490-559 37 7,6 1,9 7,3 15,9 8,1
560-629 32 6,5 1,1 5,9 14,7 7,1
630-699 19 6,9 1,4 6,2 14,9 7,4
700-769 21 6,8 1,2 6,3 14,9 7,4
770-839 17 6,4 1,0 5,6 14,3 7,0
840-909 11 5,7 0,6 4,9 13,6 6,4
Difference Celsius/100m 0,3 0,3 0,3 0,3 0,3
I got rather constantly a -0.3 °C per 100 meter elevation, so every 33 meters seems to be a difference of – 0.1 °C
re:58
Oops, I forgot to mention, I used (simply, plain ‘n stupid, as I am) the
number of the 0-69m class and subtracted the 840-909 class, divided it by
910 (0-909) and multiplied it by 100. The provided data had one digit after
the decimal point/comma, so I limited the result too, to one digit after the
decimal point/comma, as is should be done.
Re #s 30, 36 and 42
Has anybody noticed in the referenced site,
Click to access i1520-0477-72-11-1718.pdf
that the photo in Figure 2 is a phony? It is a composite of two pictures, one of the CRS (A) on the left, and the other of the MMTS (B) on the right, which appears to have been patched together. Note that the caption is a little evasive on this, simply stating that the photograph shows the instruments “photographed at the same scale” but not that two separate photos were spliced together. The possibly deceptive image is clearly seen in the shadows of the right-hand legs of the CRS figure which are missing where the MMTS picture was overlain. The site appears to be a flat-top roof. Why go to the trouble of lining up the two separate pictures in a way to make it appear that it was one picture?
A possible answer is that, as the paper describes, the comparative measurements were not taken at the same time. In fact there is a deliberately built-in 5-month lag between the times of the CRS measurements and the subsequent MMTS data used in the analysis. There is no indication in the article that any of the measurements were ever actually taken simultaneously at the same site, but rather that the coincident readings were provided by nearby “representative” sites that spanned both periods and adjusted for the bias differences determined by this comparison.
I’m interested in comments from others here on how valid such an analysis might be. One would think that they could have at done at least a few side-by-side measurements from sites where both instruments might have been making measurements at the same time, if for no other reason than to check the validity of their reconstruction.
An innocent explanation might be that they didn’t recognize the need to do such an on-site validation until after the replacements were well underway.
Anthony, anybody:
Somewhat OT but a friend and I surveyed a couple of sites today in rural Georgia, our first. One station had a MMTS and both a Standard Rain Gauge and a Fisher-Porter type precip gauge. The F&P was located under a tall bush or small dense tree and the “funnel” entrance appeared to be clogged with leaves from a branch 1 foot above. There was standing water (2-3 inches) in the funnel.
I do not know these gauges (had to look up what they were). Can I assume that it is indeed clogged and non-functional?
RE61, yes, non functional for sure if the funnel is clogged…it will read less rain than occurs, though some may trickle through.
#38,
No, that just makes Anthony a shill for big Whitewash and all his data and testing irrelevant 🙂
#60 aurbo,
I looked at the picture and it seems like an attempt at a deception. The caption indicates the picture was taken at the same scale. It looks also like an effort was made to replicate the time of day (shadow angles).
A more honest picture would have the demarcation line between the two photos emphasized.
Re #65 (and with apologies to Sir Walter Scott)
One wonders if this is where the weaving of the web began. The data quality issue sure is a tangled mess today.
#60, 65, 66 etc.
Its Fig 2(a) and Fig 2 (b), depicting two different types of temperature monitors. The caption needn’t specify they are diffrent photos, as that is implicit in the description. You guys are jumping at shadows (seriously).
RE67, I’ll have to second Alan’s opinion. yes the photo is a composite, but I think it done for the necessity of preserving column space in BAMS, and not for any deceptive purpose.
#67 Anthony,
I repeat. A more honest photo would have emphasized the demarcation line. No extra space required.
An honest mistake? Could be. I am sceptical.
Re#67.
Its neither dishonest nor an honest mistake. They’ve simply attempted to make the figure aesthetically pleasing. You’d be surpirsed how long people will fiddle with a figure to make it look ‘right’. I simply can’t see how that figure implies anything about the data. Its simply portraying two different temperature measuring thingys. You might say to that, well, why bother then? Answer is that most people try to get as many figures in their papers possible. Sort of like the ‘a picture tells a thousand words’ theory.
#49 Wolfgang:
You generally expect solid surfaces at night to be cooler than the air, because they radiate away heat better than the air does. That is why you can get frost on the ground when the air temperature stays several degrees above freezing. The freeze protection logic in my pool controller automatically turns on the pumps when the temperature falls below 40F (4.5C) for this reason. Solar water heating panels generally automatically drain at about this temperature for the same reason.
I remember from my college heat transfer courses that a clear nighttime sky can be treated as having a -40C temperature for radiation-exchange purposes.
The interesting thing for me is that all three surface types seem to have basically the same emissivity in the far infrared as indicated by their virtually identical nighttime temperatures. The difference in daytime heating between whitewash and latex indicates a significant difference in absorbtivity/emissivity (which must be the same for a given wavelength) in the near infrared — Anthony’s experiment so far seems to be a confirmation of the posited difference here.
#70 Alan,
How a bout bad captioning:
A side by side photographic comparison…
BTW thermistors are notorious for their calibration problems over wide temperature ranges.
wiki:
http://en.wikipedia.org/wiki/Thermistor
For accurate temperature measurements, the resistance/temperature curve of the device must be described in more detail. The Steinhart-Hart equation is a widely used third-order approximation:
1/T = a + b * ln[R]+ c *ln^3[R]
Some typical values for a b c
a 1.40e-3
b 2.37e-4
c 9.90e-8
from: http://www-odp.tamu.edu/publications/204_SR/112/112_t2.htm
A 1.134754819E’€”03 1.13442892E’€”03
B 2.333425717E’€”04 2.33373683E’€”04
C 9.032046858E’€”08 8.99784317E’€”08
In other words every thermistor needs its own calibration if it is to be accurate (.02 C is possible) over wide ranges.
If you don’t want to use the Steinhart-Hart equation you can do a table driven read out with linear interpolation which is a fair approximation over narrow ranges.
#60, 69
You have got to be kidding! A more honest photo!??? Someone trying to be dishonest??? The point of the picture was to compare the two objects not to somehow convince anybody that the two objects are inappropriately situated next to each other. How often do you take a single photograph of two people to show them together and then say “By the way, they are photographed at the same scale!” If you want to criticize the content, then at least do it on something of substance.
RomanM
I’m going to agree with Roman.
Merely by stating that the two objects are photographed to the same scale, he is declaring that there are two photographs in the picture.
If it were one picture of two objects side by side, there would be no need to mention that they are scaled the same.
#60 et al.
I agree with Anthony that the composite most likely is not deliberate deception. In the biological sciences literature images are frequently aggregated this way to save space and everybody recognizes that they are not pictures of cells lying side by side on the same microscope slide.
Re #s 60, 65, 66, 68, etc…
I said “deceptive”, not “dishonest”. The caption is presumably accurate, but without a careful parsing of what is, is, one might be left with the impression that the instruments were side-by-side, or at least both present when the observations were recorded.
The real question is the validfy of the methodology used to compare the data by reconstruction. And, how does one replicate the analysis in the absence of any of the rawdata or a solid reference to it?
aurbo, I agree with you that there are questions on just how valid their statistical techniques were. They describe “pairing” 424 MMTS stations with 675 CRS stations using correlations. For one of their analyses “a weighted mean of the five most highly correlated CRS stations (weighted by r) was paired with the MMTS station”. Weighting “by r” is a new one to me. And simple arithmetic tells you just how often each of the CRS stations needs to appear “paired” with an MMTS station. Then they used simple t-tests (which assume independent values) to compare the differences and to calculate error bounds. The claim is made that the analysis was robust because “a very large sample of data was analyzed!”
A bit of cosultation with a stistician might have told them they would get better results by leaving the older CRS devices in place alongside a new MMTS device at perhaps 100 suitably selected sites (just like in the composite picture), for at least a year or two.
By the way, why do they (and other people involved in temperature studies insist on calling the “mid-range” a “mean?”
RomanM
37,
As stated above, solid objects will gain and lose heat faster than the air around them because they have higher absorptivity and emissivity than humid air. So the “higher highs and lower lows” would be expected to be true for virtually any solid object when compared to air.
47,
Air temperature generally stops declining when it reaches the dew point. The solid objects don’t have the same constraint, but will be buffered by condensation and the humidity of the air around them. So as the air T declines and the humidity approaches the dew point, the decline in T will level off. It appears from Anthony’s data that the T of the wood during declining T does not exceed about 3-4F below the air temp, which is probably related to the buffering of condensation. When the air temperature levels off at or near the dew point, the solid object levels off as well…about 3-4F below the air temp.
One thought on all this is whether some of the older stations might have had a Lead based Paint applied and what its effect might have been.
Also in order to address the issue of the black plastic and its different radiative and absorption characteristics would it be worthwhile to cover the stands with a White sheet?
Just some thoughts.
RE 72, not only that, thermistors change their response with age. And since there is no periodic calibration of the MMTS system, only replacements, they have no idea what the error of the data is.
The MMTS does have a calibration plug, basically a precision resistor in a plug, which is inserted into the base unit to see if the display reads correctly, but there is no method of field calibration for the sensors. The plug only calibrates the display for one temperature, 77.1 degrees F. The remainder of the range response is unknown. The MMTS spec calls for only 1 degree of accuracy, with 0.1 degree F resolution.
http://www.srh.noaa.gov/ohx/dad/coop/EQUIPMENT.pdf (see the section on MMTS calibration)
So the signal that is being sought in the surface temperature record for climate change over the past century is masked by the inherent noise of the MMTS instrument, not to mention all the other UHI and microsite effects we’ve seen. The magnitude of the signal being sought is about the same or less than the cumulative error noise. Pulling useful data from this USHCN surface network is looking less certain each day.
I equate pulling a climate change signal from the surface temperature record to trying to listen to a speech via a distant AM radio station through heavy static. You may be able to catch bits and pieces, but the whole message is not complete.
If I were to write a paper, as Dr. Pielke suggested, using the NWS MMTS system to measure CRS response to paint rather than the NIST calibrated system I’ve used, I’d be laughed at because the instrumentation accuracy was far too coarse to ascertain the signal.
RE79, I’m sure that is the case, as lead based paints were the norm for a few decades. Getting lead based paint today to run a test will be a tall order. But its in the mix of tests I hope to be able to do. I’m also hoping to get some paint samples from older screens to have them examined by a spectrometer.
#71 Curt,
thank you. I understand I’m still intuitively mislead by the inappropriate term ‘air temperature measurement’ while what’s actually monitored is the result from a mixture of convective (and latent) heat transfer plus/minus radiation balance.
Might be a nice additional experiment for Anthony: A comparative more ‘real air temperature measurement’ by forced convection. E.g. a small fan in a white tube with a temperature sensor installed close to the intake.
#82
REM:
Here are some links about night sky temperature aso:
about the sky temperature
The effective sky temperature
MEASUREMENT OF NIGHT SKY EMISSIVITY IN DETERMINING RADIANT COOLING FROM COOL STORAGE ROOFS AND ROOF PONDS
Temperature of a Glass Plate Exposed to the Night Sky
Heat Transfer Model for concrete pavements and bridge decks
Anthony, maybe you should ask the NWS for some old Stevenson screens. E.g. teh one at Fallon that’s being replaced.
#81 Anthony, you should be able to get lead-based paint from marine supply stores.
http://www.kirbypaint.com/products/
RE85, Thanks Howard, It looks like I’m going to have to run another paint experiment, using whitewash, oil based, lead based, latex, and enamel paints…and of course, silver “rust-o-leum” from the Leech Creek Dam USHCN station fame.
The point here is that there is no record of painting these cotton region shelters in the station metadata, other than a casual mention. Determining what sorts of effects paints that were commonly available over the last century will give a clue towards assessing the magnitude of the problem.
Like I said earlier in this thread, there is the possibility that when I do the full scale tests on the screens, the amplitude of the effect may be minuscule compared to the effects seen in wood borehole test data posted in this thread.
This is my first post, I am not a scientist, just have a little engineering background. The measured temperature error is interesting. It explains some of the confusion I have had the hockey stick. Have you considered following the money. The weather derivation futures are big money. I would be curious how futures trades sort sites for reference. It could also be a good source of donations to complete your research.
Dallas Tisdale
#80 Antony,
Thanks for answering a question I have been asking for over a week re: calibration. There is no repeated recalibration.
The next step is to find out what kind of algorithm was used in the electronics. Plus initial calibration methods.
The resistor test assumes that all the thermistors have identical curves within the specified error limits. It ain’t necessarily so. Then you have the aging problem.
Anthony:
I see from the photo you are not watering the lawn. Good for the experiment, but bad for the grass. Try to make sure all of the enclosues are over equally dead grass, which should be easy.
I did some browsing regarding heat transfer through wood. There may be another issue with the Stevenson boxes. Heat conductivity is increased as the density of the material decreases and the moisture content increases.
See http://www.fpl.fs.fed.us/documnts/pdf2007/fpl_2007_gu001.pdf for computer modeling of softwoods.
As the wood of the Stevenson screens ages, joints will open, the wood becomes more exposed and the moisture contect of the wood increases during wet seasons. This creates a potential for wood to rot and deteriorate the cell structure making the wood less dense (especially near joints and ends). Therefore as time progresses, the moisture content of the wood increases over time and the density decreases. This will affect heat transfer in wet cold seasons due to moisture content and dry warm seasons when there is less fiber to act as insulator.
The magnitude of these effects will vary dramatically from none in newer boxes to significant in older rotted boxes. The actual effect on the average readings may be insignificant, but it is something to consider.
An interesting computer program called “MOIST” is used in this study http://www.fire.nist.gov/bfrlpubs/build95/PDF/b95042.pdf that may be useful for your own testing.
We engineers use a rather primitive testing technique to determine wood rot: we stick an ice pick in the wood. If the pick meets a lot of resistance and barely penetrates, the fibers are in good shape. If the pick penetrates easily and deeply, there has been some rotting and fiber damage.
I can see it now when we’re all done….
“Let’s spend billions of dollars putting frozen CO2 into the ground, the paint on a bunch of thermometer shelters is getting warmer!”
Anthony:
You may want to capture cloudiness and windiness data from a nearby weather station. These variables may affect your results.
Done, see the wx_data file in the thread above. It has wind and solar radiation
SteveM and Anthony,
It might be interesting if one could find sites that have continued to use the CRS..
It might be interesting to list sites that havent moved or have few moves
Surfacestations.org made drudgereport.com this morning. It is short on bandwidth and no longer responding.
I wonder if the whitewash/latex controversy may be resolved by comparing wet/dry temperature readings? As I see it the whitewash is porous, and the latex is impervious to water thus the whitewashed screens may be affected by humidity and cooler due to evaporation of absorbed water:- thus older records would respond to damp screens. Modern painted screens would degrade in time if not repainted, but if remaining dry would suffer from extra irradiated heat inside. So, the wet/dry difference may be lower in whitewash, higher in latex.
Anthony,
Do you have anything new to report on your study of Stevenson screens, whitewash, latex paint, etc..?
One Trackback
[…] Audit has a very interesting posting on the role of paint on measured temperatures that ia being investigated by Anthony Watts. This is clearly a publishable result when completed and Climate Science urges Anthony to do […]