The Stevenson Screen Paint Test

3 Stevenson Screens ready for paint test

You may recall that back in May I did a simple preliminary experiment to give me guidance on a hypothesis: That changes in paint on Stevenson Screens over time make a measurable difference on the temperatures recorded inside them. This stems from the fact that when the Weather Bureau commissioned the design in 1892, whitewash was specified. But whitewash is no longer commonly available, and the National Weather Service changed the specification in 1979 to be semi-gloss latex paint.

But, cured whitewash is composed of Calcium Carbonate, while latex paint uses Titanium Dioxide as a pigment. While they both appear “white” in visible light, they have vastly different properties in infrared.

My first simple experiment used thermistors in boreholes into 3 wood slats; 1 bare wood as a control, the other two painted with whitewash and latex, showed me that there was a measurable difference in the temperature of the wood by as much as 2-4 degrees at times. I needed to do that experiment before I embarked on the full scale test, since each of the Stevenson Screens you see here in the pictures cost me about $1000.00 Since I’m doing this out of pocket, with no funding or grants, I had to try a small scale test first.

The photos show 3 standard Stevenson Screens as used today in the United States. One is bare wood, unpainted, as a control, the middle one is latex, as sent from the supplier, and the third is painted with a historically accurate (for early 20th century) whitewash mixture that I obtained both materials and formula from the head chemist at the National Lime Company.


The device on the tripod is a stacked plate IR shield with a small fan to pull air through, commonly called an aspirated shield. It is the air temperature reference and placed at the same exposure height as the thermistiors in the screens. Also nearby but not shown is a pyranometer to measure solar insolation and wind speed/direction sensors that are being datalogged as well.

Each Stevenson Screen and the air temperature reference sensor are fitted with matched, calibrated thermistors, NIST traceable with certificates, that are connected to a calibrated data-logger, also with a certificate. The resolution is .01 degree Fahrenheit with an accuracy of +/- 0.1 degree over the range.

I expect that the air temperature differences inside the screens will be less than the 2-4 degrees I observed in the paint slat test. It’s possible that there will be no significant difference at all. I won”t know until I run about a months worth of datalogging.

The site, while not ideal due to the trees, is the best I could get permission to use. Fortunately the trees do not directly shade the screens except for a short portion of the day. It’s also out of the way, so vandalism will not be likely. Since it had to be an unwatered grass field, concerns over fire danger were raised from some I asked because of the electronics package, so I had limited choices. Perhaps later I’ll be able to find a better site but for now it will have to do.

The whitewash on the third Stevenson Screen is still curing, as the chemical reaction is not yet complete to convert Calcium Hydroxide to Calcium Carbonate. In about a week, I’ll make the data available via a web link in near real-time.

0 0 votes
Article Rating
Newest Most Voted
Inline Feedbacks
View all comments
July 14, 2007 7:12 am

The unpainted screen appears to have some grass growing around it. Could this distort the result?

July 14, 2007 7:52 am

How are you going to get accurate readings without an air conditioning unit nearby?

July 14, 2007 7:58 am

Regarding the shade… Which box gets the most?
Based on your description I’d guess the two edge boxes (control, historical) are most likely to get it, not the latex box. How do you plan to account for this discrepancy in your data analysis?

July 14, 2007 8:31 am

Ken, Thanks for bringing up the question.
I have two methods:
1- Next week I plan to do some sun angle/treeline measurements, so that I can determine a window within which no shade is received. Because the main driver of the experiment is solar insolation, and because the data is timestamped, I can then exclude data outside that window. IE just look at “full sun” and “full dark” data.
2- you may notice that the screens are not anchored to the soil. The reason for that is so that I can swap positions and run the test again. I also plan to run a test where all three screens are over a concrete slab, gravel, asphalt, etc, since we’ve seen that sort of siting issue. Getting data on how the screens perform over different surfaces is equally important.
Bishop: “The unpainted screen appears to have some grass growing around it. Could this distort the result?”
Indeed it would, but its starthistle, not grass, and it’s density is low. I have permission to bring in a weed wacker next week to take them down.
The actual experiment won’t start for a week, since it will take that long for the whitewash to cure, so if anyone sees issues that need addressing, don’t be afraid to post them.
While I admit the site is not perfect, I had to choose between this site or not doing the experiment at all. I did have a perfect site originally, but the owner had concerns about fire danger that I could not assuage, even though I explained the thermistor cables carry only tiny voltages and currents.

July 14, 2007 8:46 am

I should mention that in my quest to ensure accuracy of measurement, I did in fact install a small a/c unit nearby. But it’s not what you think.
By necessity, the data-logger has to be outside, and it is 75 feet away from the screens. I wanted 100 feet, but I could not get calibrated probes with 100 foot cables, so I had to accept what was available.
The data-logger would be exposed to sun, dust, and diurnally varying temperatures. I was also concerned that over 100 degree temperatures might fry the electronics, since they have a 7805 (Lon I’m sure you know the part)voltage regulator on the circuit board with no heatsink.
To remove that worry and potential accuracy problem, I placed the data-logger in one of those 12 volt DC powered “camping coolers” to ensure a constant temperature, shield from sun, and insurance against 110 degree afternoons.

July 14, 2007 1:51 pm

Great work, Anthony!
As for a shortage of “whitewash”… I can certainly see why.
My gosh, look at how much they need to use in Washington and in Sacramento on a daily basis!
Heck, they’re gunna need thousands of gallons of the stuff to try to cover up :- )

David Walton
July 14, 2007 2:07 pm

Hey, nice set up, Anthony. It looks like you snapped the photo around noon and to me it doesn’t look like shade will be a factor (except in winter, perhaps). Which way is north?
If you need a hand with weeds I can always use a little exercise.
I also have a little Round Up, if that is acceptable.

Mark Francis
July 14, 2007 5:21 pm

Anthony: Keep up the great work!

Soren Monies
July 15, 2007 1:43 am

It may not apply in your setup, but the cabling from the screens could be subject to eh, datamanipulation from the critters living in the undergrowth…? One test-setup I have information about, had to route the cabling through hydraulic hoses (steel armored) to protect from the gnawing little beasts

July 15, 2007 5:24 am

You are far to busy to go looking round for my site so here is a copy of my post today:
“The pursuit of knowledge through scientific enquiry has a rather romantic, even glamorous, image. One imagines all those famous “Eureka!” moments. Well, one has to imagine them because mostly they never happened! However, for those prepared to spend a few minutes looking at the decidedly unglamourous face of practical scientific enquiry, you will do no better than visiting here. The indefatigable Anthony Watts has done us all a service by mounting a national campaign to photograph the climate temperature sites in the USA from which much of the data is used to support ‘Gormless’ Gore and his ilk. That is a story in itself but Watts has taken it one stage further.
All official US temperature recording equipment, at their various sites, is housed in specially designed timber boxes called Stevenson screens and have been so since 1892. Originally the specification for these ‘boxes’ was that they should be painted in whitewash. However, ‘Watts the Diligent’ decided to investigate and found that “… the National Weather Service changed the specification in 1979 to be semi-gloss latex paint.” Why does this matter? Because, as Watts points out, “cured whitewash is composed of Calcium Carbonate, while latex paint uses Titanium Dioxide as a pigment. While they both appear “white” in visible light, they have vastly different properties in infrared.” A very quick and not to be relied on test indicated a difference in temperature between the differently-finished timbers of 2 – 4 degrees at various times.
That was enough for ‘Watts the Seeker of Truth’ to fork out $3,000 of his own money to buy three Stevenson screens to test his hypothesis. One is left unpainted to act as a control and the others are painted respectively in whitewash and semi-gloss paint. He has gone to considerable trouble to pick a site as clear of interference as is possible, in direct contrast to the existing official sites! We now await results – but the results are not the prime importance here, it is the careful, unglamourous but dedicated pursuit of scientific knowledge that Watts exemplifies that is wholly deserving of our admiration and, given the importance of the field in which he is working, our gratitude.

July 16, 2007 2:39 am
Jim Brenner
July 25, 2007 11:53 am

Here in Canada, it was reccomended that the screen be mounted 4 ft. above a grassed in surface, and that the grassed be kept mowed below 6″. Not an easy task, however it is possible using hedge shears.
The standard stand used was 2×4 cedar, all surfaces, including the screen, were painted with a good quality exteriour enamel.
The principle behind the Stevenson Screen is to provide a standardized method of measuring exteriour air temperature. If indeed the colour of the surface affects temp. readings, that tells me that it`s time for a new design..

John Hamon
July 25, 2007 4:05 pm

Dear Jim from Canada:
I live about 50 miles north of the 49th. I’ve crossed into Montana at Opheim and other stations. Many US Customs offices had a Stevenson screen, along with other instruments, just a few meters behind their station.
I recall discussing this topic with a U.S customs officer who had recorded temp. for many years. I think the grassed-in and well trimmed surface is extemely important. I also believe that the traditional stand (cedar) has now been replaced with metal. Heat absorption varies greatly between metallic surfaces. I’d trust the cedar painted with semi-gloss.
Environment Canada has now replaced most stevenson screens with plastic or vinyl imitations. Temp., rainfall, snow etc. are now recorded by computer and the data is transmitted via satelite. Many weather stations are isolated, away from populated areas. It greatly lacks the “human touch” of a manual recording. Anthony is working on a worthwhile project!!

David Clark
August 18, 2007 9:47 pm

How is the the screen paint test going? Do you have any idea how those little louvered cylinders we see in so many of the surface station photographs compare to the Stevenson screens in terms of temperature influence?

September 9, 2007 5:39 am

And? How is the test going?

Anthony Watts
September 9, 2007 9:37 am

The first months worth of continuous data is now completed. I’ll analyse and post it soon.

October 1, 2008 5:48 am

Hi – any dates yet on when we can expect to see the results

%d bloggers like this: