New paper finds that aging weather stations record much higher daytime temperatures, 1.63°C higher than new stations
While we are all watching the heat wave developing in the US southwest, here is something to consider. Albedo on the surfaces of weather station shelters changes with time, something I and the volunteers have documented with the Surface Stations project. For example, here’s a aged weather station where the whitewash coating is coming off and the bare wood is becoming exposed:
Back in 2007, Pat Michaels wrote in an American Spectator column “Not so Hot“:
“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.”
Now there is proof that changes in station shield surfaces affect temperature
A paper published in the International Journal of Climatology finds that aging of the solar radiation screens on weather stations is causing a large positive bias in measured temperatures of 1.63°C, which by way of comparison is more than twice the global warming of 0.7°C recorded since the end of the Little Ice Age in 1850.
According to the authors, “During the comparison [of the new vs. 5 year old] and 1 to 3-year-old screens, significant temperature differences were recorded at different times of the day. The differences, wider than the uncertainty amplitude, demonstrate a systematic effect. The temperature measured with the older screen is larger, and the maximum instantaneous difference was 1.63 °C (for 0–5 years comparison) in daytime hours.
During night-time the two AWS’s measure the same temperature (within the uncertainty amplitude). This behaviour, increasing with increasing solar radiation intensity and decreasing with increasing wind speed, is attributed to a radiative heating effect. The screen ageing has compromised the shield effectiveness introducing a significant change in the temperature evaluation.” The paper is yet another blow to the unreliable, biased, and highly upward-adjusted temperature record.
The paper: http://onlinelibrary.wiley.com/doi/10.1002/joc.3765/abstract
Comparative analysis of the influence of solar radiation screen aging on temperature measurements by means of weather stations DOI: 10.1002/joc.3765
G. Lopardo et al
Abstract:
Solar radiation screens play a key role in automatic weather stations (AWS) performances. In this work, screen ageing effects on temperature measurements are examined. Paired temperature observations, traceable to national standards and with a well-defined uncertainty budget, were performed employing two naturally ventilated weather stations equipped with identical sensors and different only for their working time. Three different tests were carried out employing different aged AWSs: a 5-year-old AWS (AWS5) was compared with a new device (AWS0), a 1 year old (AWS1) was compared with both a 3 years old (AWS3) and a new one devices (AWS00). Due to solar and weather conditions exposure a degradation of the screen reflective coating is evident for the older AWSs (5 and 3 years old) and so a qualitative estimation of how different conditions of ageing affect the temperature drift was done.
During the comparison 0 to 5 and 1 to 3-year-old screens, significant temperature differences were recorded at different times of the day. The differences, wider than the uncertainty amplitude, demonstrate a systematic effect. The temperature measured with the older screen is larger, and the maximum instantaneous difference was 1.63 °C (for 0–5 years comparison) in daytime hours. During night-time the two AWS’s measure the same temperature (within the uncertainty amplitude). This behaviour, increasing with increasing solar radiation intensity and decreasing with increasing wind speed, is attributed to a radiative heating effect. The screen ageing has compromised the shield effectiveness introducing a significant change in the temperature evaluation.
The experimental results of a further comparison, between 0- and 1-year-old screens, confirm the same conclusion showing a negligible ageing effect, within the uncertainty amplitude.
h/t to The Hockey Schtick
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
So I’m sure someone will come on here and claim that homogenization practices should correct for this through their statistical methods. I say probably not…and almost certainly not with 100% success. There statistical methods are looking for high frequency step changes…nearly of the instantaneous variety…right? So if someone cuts down a tree that was previously right next to the sensor, if they erect a building right next to sensor, change paint colors of the neighboring building, install new A/C next to the sensor, move the station, etc–then homogenization might work. It should also work to filter out instantaneous negative biases like the installation of a sprinkler system.
But homogenization will often fail (or even work in reverse of how it should) when experiencing slow/gradual interferences to their measurements. This is because the bias being introduced on the signal is in the same frequency range as the signal of interest–and therefore the range being passed by he homogenization “filter”. Thus, something like UHI can get in easily. A new building right next to the sensor–easy to filter out since it may cause a 1 C change in the signal that goes from zero to the full effect in a couple of months. But what about a building 3 blocks away that only introduces a 0.05 C effect, for instance? Doesn’t seem like much until one of those buildings was put in every other year for 20 years…a full 0.5 C increase at that point…added in nearly a linear fashion
So how does homogenization treat something like what we’re seeing with the aging weather stations? Here’s my hypothesis…the aging effect is a slow change in the signal and for practical purposes a continuous function (the exception being something like a large storm causing a step-like function if it severely damaged the paint). Thus, we get a warming bias in the signal that isn’t filtered out–because it’s a smooth, continuous function. This bias is present, to some degree, in every single station experiencing weathering.
A scenario: one day, after ~20 yr of operation, people tending a station decide to put fresh paint on the station. This gets the raw data back to unbiased, but since it happens as a step change, the homogenization process sees it as a problem (“station change”) and “corrects” it back to be in line with all the other neighboring stations that have been experiencing the same creeping bias! Worse than that, a few years later one of the neighbor stations that was used to justify the “correction” of the raw data will be repainted, seen to experience a step change, and then its raw data will be “corrected” back up to be in line with the others that are experience a slow warming bias! And the first station in the scenario, already “adjusted”, will be used as part of the justification for this “correction” because it will have resumed the weathering process after being painted fresh.
Now, all that said, I’ve not played with the statistical methods employed for the homogenization processes, so it’s possible that my scenario given above couldn’t play out. If so, will someone please explain to me how situations like the above could not affect the results with a bias?
Additionally, I’d say the above could actually be tested with real station data. Simple – do a one-day paint job on a station and see how the homogenization process modifies its raw data both before and after the paint job. At a different location but one with similar temperatures/insolation, paint the station in 1% increments once per week so that it takes 2 years to replace the surface. One might want to keep applying fresh coats to the already-repainted sessions while doing this. Check out how the homogenization process handles that type of data change. Is it not “corrected” in the same way or even at all? Would be an interesting test.
-Scott
This is truly terrifying. That weather station looks an awful lot like an old farmhouse that some friends and I rented back when we were in our twenties. So the POTUS and the other so-called leaders of the western world are going subject several hundred million people to policies based on readings from instruments housed structures that are identical to the dilapidated structure a bunch of twenty-somethings partied in 30 years ago.
Hope and change!
I think my method to measure day time rise and following night drop rejects these types of errors(follow the link in my name for paper). Interestingly i dont find a trend in warming.
Anthony, with your site survey, if you or someone can provide some station numbers, I’ll extract the data, maybe some good and bad stations?
There are two courses of action that should be taken here:
1) incessant demands that GISS be adjusted for this effect
2) a volunteer project to paint stations. Seriously. You too can help fight global warming!
Probably the newest Stevenson screens on the planet are the ones they sold Anthony six years ago. Don’t that bring back memories, huh?
🙂
Andy, Tilo Reber and Scott: I participated in the same discussion with Zeke at the Blackboard. Zeke’s breakpoint detection algorithms find artifacts that result in a sudden cooling (going forward in time) with respect to neighboring stations more than once a decade in the US record. These cooling artifacts are far more common than warming artifacts, and result in corrections that lower early temperatures. In that discussion, I asked if occasional station maintenance might be responsible for a sudden correction of gradually deteriorating observing conditions that restored initial or normal operating conditions. Correcting a breakpoint caused by slow deterioration followed by restoration of earlier operating conditions would introduce a bias in the record. Slow gradual deterioration might be missed by algorithms whereas artifacts associated with maintenance would be much easier to find.
A slow change screen albedo followed by painting or replacement would fit this degradation/maintenance scenario perfectly. Breakpoints caused by gradual deterioration of station albedo followed by painting or replacement would be easy to diagnose, because they would be found only in the maximum temperature record (and average temperature record), but not in the minimum. Major breakpoints associated would a station move would likely be found in both records.
Frank says:
June 29, 2013 at 9:12 pm
Hi Frank,
Thanks for the info in your comment. Any idea whether the current algorithm considers the difference in max & min behavior as the paragraph I quote from you suggest should be possible to do?
Thanks,
-Scott
This paper points out the danger in using a scalpel on long term temperature records.
From Prior discussion in: Berkeley Earth finally makes peer review – in a never before seen journal – Jan 19, 2013 WUWT
It is a reasonable hypothesis that if a Stevenson Screen is painted, every 5 years for example, there is a gradual instrument drift between paintings. Then on the day of repainting, there is a sudden shift, a recalibration, of the sensor. With temperature changes on the order of 1 deg C on the repainting of the Stevenson screen, is it unreasonable to suspect that Best will preferentially slice at these recalibration discontinuities. If so, then BEST is making a biased record worse, by repeatedly incorporating instrument drift from degrading Stevenson Screen albedo as natural warming and discarding the recalibration resets that are the act of painting.
How do the well sighted rural stations compare with the new very high tech stations?
If they always used the same type of paint, and they painted the same number each year, it might all wash away in the averaging. But suppose that the repainting rate has stayed the same but the paints have improved with time? Then modern numbers will be more accurate while the ones from long ago will tend to be too high. Oooooops! Better adjust those old numbers downward! Voila, another justification for concluding temperatures are still rising.
So repainting a thermometer shelter may trick the GISS adjustment process into adjusting the station upwards because the aging process is slow and unnoticeable while the repainting event introduces one-time sudden temperature drop.
You get positive bias if you don’t repaint. And you get positive bias if you do.
Things are starting to make sense.
A government testing lab where I worked in the 1970s was responsible for maintaining a number of Stevenson Screens, distributed across the state of NSW (Australia). Those enclosures were designed to use structural fibrous cement in the body of the cabinet and not laminated timbers. The screens did not require painting, but were progressively replaced by non-asbestos screens to comply with Australian standards by the late 1980s (or completely removed from service).
I wonder how extensively such non-painted materials have been used elsewhere.
Here is a restating of my Jan 23, 2013 11:30am comment concerning on how the BEST scalpel is the wrong tool to use in a time series populated by recalibration events such as Stevenson Screen repaintings.
In any arbitrary temperature record let us identify time periods
A0……A9.B0…….B9.C0…….C9.. etc.
The spans A0….A9, B0…..B9, C0…..C9 are run periods of a sensor with instrument drift, such as a degrading paint coat on a Stevenson screen.
A9.B0, B9.C0 are recalibration events where the Stevenson Screen gets repainted, electronic thermometers are recalibrated or replaced and other maintenance gets replaced.
In this arbitrary temperature record, what are the time periods with the most accurate temperatures? It must be A0, B0, and C0, the times when the technician has packed tools and leaves.
Worst case, a delibrate cherry pick, a boundary condition,
the time series A0……A9.B0…….B9.C0…….C9.D0….. is a saw-tooth wave of gradual instrument drift, followed by a discontinuous recalibration. The BEST scalpel process will make cuts between A9 and B0, between B9 and C0. The BEST scalpel will keep the instrument drift and discard the recalibration. Drift (error) becomes signal and the periodic recalibrations (restoration of accuracy) are treaded as noise to be ignored.
The recalibrations are vital to the long term records and must not be ignored.
Hansen described satellite temperature measurements of Earth as “obviously wrong” !
It’s far more likely that satellites are obviously right.
I guess the adjustments needed to account for the age of the housings plus UHI would probably bring the temperature series more into agreement with CONUS and the satellites.
As soon as NOAA paints their weather stations, global warming may be over ?
The subject of the following Dutch report (http://www.brusselnieuws.be/artikel/stad-tot-5-graden-warmer-tijdens-zomernacht) is not the albedo on the surfaces of weather station shelters but UHI. According to recent measurements of the heat island effect in Brussels, in daytime hours it is 1 to 1.5 C degrees warmer than outside the city and even 3 to 5 C degrees warmer during night-time!
According to the official measurements, the temperature in Uccle, a municipality being part of Brussels, the temperature rose with about 1°C per century. So the question arises whether the temperature increase with one degree in Uccle, (which is larger than the global temperature rise (about 0.6 to 0.7 degrees)), is due, in whole or in part, to the heat island effect. In 1850, the weather station was located in an area with few buildings while Uccle now has become a densely populated town with only in the southeast the cooling effect of a forest.
One can also ascertain that the largest impact of UHI seems to happen during night-time while the impact of the albedo of the paint on weather stations is largest in daytime hours.
Looks like the screens painting stopped mid 1990s.
Of course, this will not necessarily enhance the observed warming over any particluar period. The key question is how has the average condition of weather stations changed over time (and how is it changing, going forward in time). Some stations have doubtless almost always been in poor condition. My guess would be that in recessionary times maintenance is pretty poor – so that the flat temperatures observed over the last decade and a half might well actually correpond to a fall in ‘real’ temperatures; though, OTOH, Anthony’s campaign may have encouraged a bit more attention to their condition (which would have reduced apparent observed temperatures!
Wow! I never cease to be amazed at all the different factors that we assume are constant and then find out they need to be considered and normalized.
Tree roots, paint jobs, precipitation, shear in glaciers – all kinds of stuff I’ve learned about just in the last few months. And then there is the Divergence Problem, too. And clouds. And Milankovich Cycles. And comic rays. And on and on. . . But especially when Phil Jones can’t do a simple trend line in Excel. LOL
Oops! Comic rays. ROFL. . .
Reminds me of that old comedian (LONG ago) with the bit about “You can call me Ray, or you can call me Jay, but ya doesn’t has to call me Johnson!”
Screens made of sheets of pressed steel which are coated with white enamel should resist weathering effects for decades.
But, of course, they will cost you dearly…
@Rik Gheysens June 30, 2013 at 12:40 am:
“In 1850, the weather station was located in an area with few buildings while Uccle now has become a densely populated town with only in the southeast the cooling effect of a forest.”
Rik – Every time I see NW European weather systems on satellite images, they are coming in from the west or northwest, so a southeast forest would be on the lee side of the city and wouldn’t seem to have any cooling effect. Correct me if I am wrong.
@J Martin June 30, 2013 at 12:13 am:
“Hansen described satellite temperature measurements of Earth as “obviously wrong” !
It’s far more likely that satellites are obviously right.”
Right, Hansen. That is why the satellites and radiosonde balloons have agreed since 1979.
(This was one of the first things that perked me up about the global warming issue… Which basically came down to: If two out of three methods of measuring the global temperature are in agreement, why is it that the two are considered wrong and the third is taken as gospel?)
Hello Mr Measurement Error, could I introduce you to Mr Climate Science? I don’t think you two have ever met!
Satellite temperatures are far more accurate than surface instrumental global temperatures, the technology is far superior. Even if they both had the same error we are comparing a complete coverage of the surface area where it doesn’t miss parts of the poles with < 1% of entire global coverage. Yet still if the satellite was <1% surface area it would still be more accurate. Only those in a CAGW agenda nothing to do with science think the surface instrumental data is better. The satellite temperatures have kept them more honest with surface instrument data, but still trying it on with changing data selections to support the cause.
How many of these data changes to instrumental stations are applied through to cover all history? The answer is none so they are even comparing a different data sub-set with different data sets over the long term history.None of them are comparing the same data points to previous periods, so how can we even say there has been any or even noticeable difference between the previous warming and cooling periods. Using the same Arctic based stations throughout the same period show no difference between the 1930's/1940's and recently.