NOTE: David Smith is doing experiments with the portable USB digital thermometers that are available here. This sort of experimentation is easy and inexpensive to do, and makes a great topic for a student science fair project. The results are easy to download from the USB thermometer into a PC for analysis. -Anthony
Seven Days in May
A guest post by David Smith
This is an update on recent field tests with remote thermometers (see the ”Fun with Thermometers” post for background).
My goal is to quantify, to an extent, the effects of microsite problems (pavement, buildings, trees, etc) on temperature.
In the current test one sensor (”A”) is currently in an abandoned baseball field at least two hundred feet from any paving, tree, structure, etc other than a chain-link fence:
This reasonably approximates a good-quality site, isolated from human microclimate effects.
The other sensor (”B”) has a split personality. On one side is a poorly-drained field while on the other side is an older asphalt parking lot:
When the wind blows from the north this second sensor tends to reflect the characteristics of the soggy field, while a southerly wind brings air from the parking lot.
An aerial view of the two sites (”A” and “B”) is here:
For this update I selected seven days in May in which the skies were mostly clear throughout the day and night. This should maximize any radiative effects on temperature. (Unfortunately, the site is warm, quite humid and windy this time of year, limiting the magnitude of any radiational microsite effects. But, despite this diminished magnitude there are still useful observations to be made.)
Below is a plot of the average temperature of “B” on five clear-sky days when the breeze was from the parking lot and the average of two clear-sky days when the breeze was from the soggy field. I’ve subtracted the temperature of the nearby baseball field (”A”) from these two averages so that the lines show how much warmer or cooler “B” is than “A”. I’ve also slightly smoothed the data.
All seven days were breezy, which mixes air and limits its time over the surfaces, so the effects are probably muted compared to days with less-breezy conditions:
This shows several things. One, when the wind is from the parking lot (red line), the temperature at “B” sensor is warmer than that of the baseball field, night and day. Shortly after sunrise the difference diminishes, presumably due to the higher heat capacity and thus slower warming of the asphalt vs the baseball field. As the sunny day progresses the heat content and temperature of the asphalt rises, reaching a relative peak at “B” in the late afternoon. As the sun sets and evening progresses the temperature of “B” remains elevated but to a smaller extent.
This “parking lot effect” should be noticeably greater this summer, when average windspeed and air mixing diminishes.
The effects when the wind is from the soggy field (blue line) are perhaps even more interesting. The temperature of “B” tends to be depressed vs the baseball field during daylight hours, presumably due to evaporative cooling of the soggy field. The effect is reversed a bit in the late afternoon, possibly when the dry baseball field is radiatively cooling faster than the soggy field.
The soggy field appears to be due to changes in drainage following a yearlong construction project nearby. This change in drainage and probably ground cover was subtle in nature and may have stretched over some time, something which may or may not be detected by a discontinuity algorithm. In this instance it was cooling but my conjecture is that most drainage changes are towards drying, and warming, not wetting and cooling.
These seven days in May are affirmations that it is a bad idea to have sensors in the vicinity of human-induced microsite changes. Changing drainage, repaving the parking lot, aging of the parking lot, changes in parking patterns, etc can all have an effect. The size of the effects in a given year may depend on rainfall, wind anomalies, etc, making it difficult to detect a discontinuity.
The 2 F to 3 F impact of the parking lot on average temperature, shown above, understates the parking lot impact on peak temperature. I’ve placed the recorder on 5-minute measurements and those frequent measurements detect short-term peaks during the day which are 5F to 7F higher than what the 30-minute readings detect. Once I get enough runs I’ll link their plot here.
I think there is another factor here. Winds from the north perhaps tend to be cooler than winds from the south?
Nice work Dave. Thanks. Can I assume the “more to come” refers to site C in the third picture? Is the measurement intervals one hour? — John M Reynolds
Amazing that the “professionals” who sited so many of these sites improperly did not do the same sort of experiments.
Maybe “professional” in this field means what it does when applied to women of the night.
Jeff,
While air from the north being colder on average could explain a difference in absolute temperature levels at both sites A and B when winds are from the north, why would it explain a difference in the temperature gap between them?
These are the types of experiments that should have been performed by those in the climate “science” community years ago. It is sad that basic data like this is almost nonexistent and must be performed by dedicated individuals.
Ladies and Gentlemen, lets see how GISSTEMP handles May 08 temps
http://discover.itsc.uah.edu/amsutemps/ (looks like corresponding to equal or lower than January 2008?).
My work laptop comes with vista which gives a large warm bias over XP due to the excessive power used from this awful OS to annoy the hell out of me. Maybe the greenies should promote XP as Vista leads to global warming. SIGH!!
Dunno, I was just throwing it out there. If it doesn’t apply, please ignore 😉
Great work.
Here are amateur scientists doing basic sanity checks on measurement systems that the professionals have never done.
In the 17th and 18th century, all scientists were amateurs and did basic, measure it and think, experiments: basic science.
Now, our high paid state “professional” scientists do not do the basic science and then decry the “amateur” for doing the leg-work. Which, strangely enough, shows the pro’s work to be horse feathers!.
Site “C” is actually the one of most interest to me. The parking lot tests are sort of a practice run.
When I surveyed USHCN sites I was impressed by the encroachment of trees and shrubs toward many of the MMTS sensors. The vegetation didn’t overlay the sensors but did block large portions of the sensors’ view of the sky.
Taller vegetation interferes with wind/mixing and outbound IR (both of which create a warming bias) while possibly reducing direct sunlight (which creates a cooling bias, as does transpiration). My conjecture is that somewhere outside a tree dripline the warming bias from blocked wind and IR noticeably outweigh the cooling bias from reduced sunlight and transpiration, especially on the southern side of a tree. Data from site “C” should help solve the matter.
I may be wrong, of course. It’ll be fun to see where the truth lies.
I suspect that this work of understanding site bias has been done by the professionals. That is why we know the differing “classes” of siting along with the expected bias’ of each class. That this siting bias is a known phenomena and yet ignored when choosing sites is shameful.
David,
I’m presuming that these temperature readings were taken at 5ft elevation and so are comparable. It would be useful to see a set of measurements at various heights above the ground. My conversation with a USHCN station curator revealed that there can be a 2-3 degree difference between ground level and 5 ft on calm nights at a good (rating = 2) station.
David,
Is there a satelite view available?
If your measurements show short term spikes, what do the ‘official’ sensors report? Taking the midpoint of Tmax and Tmin doesn’t work very well if they are not using a moving window or some kind of spike filter. Response time of the sensor also plays a critical role. The small USB sensors seem to react very quickly. What do the ‘real’ (or unreal) units do?
I know it would cost more, but a small wind speed/direction logger could answer some of the ‘spike’ questions.
Kudos for great work on a small budget!
When these sites are selected and the monitors installed are there any notes included in the installation to explain if the locations are NOT per specifications so that “fudge” factors must be included when recording the data?
Correction to my post: Should have written “… “fudge” factors must be included when USING” the data?”
One of the advantages of using delta temperature changes vs. specific temperature readings when detecting climate change is the removal of steady-state sensor bias. This works as long as the two readings remain in a linear range of the sensor. Of course this requires, absent climate change, that all daily variations over, say a year, would average out to zero.
As long as delta readings are used, the local environment is largely irrelevant. What’s still important though is to account for sensor environment changes. DS’ work is a good start into assessing the impact of these.
When assessing impact, the real question is “what was the TRUE temperature?” An interesting problem in itself. Short of making your own environment changes, possibly the best way is to place sensors around the one to be tested and averaged to remove local variation. I realize this isn’t particularly cheap.
Some questions:
1) Why was the ball field location picked? Do you think it represents a good average between asphalt and soggy ground?
2) If the average environment is wooded, wouldn’t location “C” be a more realistic representation of “TRUE” temperature?
3) Did you make any attempt at calibrating the USB sensors? I have a wireless outdoor temperature sensor near my house. I wasn’t fastidious about location. Over the course of a year, I’ve noticed a +5F +/- 2.5F variation between it and nearby BWI readings that doesn’t seem to correlate to anything other than nominal BWI temperature. IOW: it isn’t linear. I suspect it uses the same technology as your USB sensor. So, did you check your sensors?
So in other words, if we replace all paved roads and brick buildings with grass fields, we would need to start panicing about global cooling?
As long as delta readings are used, the local environment is largely irrelevant.
Unless, of course, the environment changes over the course of the experiment (too gradually to be snagged by inhomogeneity check), dragging the offset along with it. And the offset change winds up being conflated into the delta.
I am still seething over the USHCN-2 adjustments. They learned their lesson well from the “USHCN-1 experience”, namely that their version 1 adjustment page is one of the most quoted pieces of climate literature by skeptics.
Why?
Because they actually tell us how much they modify for each step of the adjustment procedure and how much they adjust overall.
I bet they’d yank that page pronto-Tonto, only they know it’s too late for that.
As for USHCN-2, they give outstandingly incomplete data. I could not even find a comprehensive bottom line as to how much upward temperatures have been adjusted.
And that slide show of theirs that attempts to make out like station siting doesn’t matter is extremely suspicious in its obviously selective choice of data. (They don’t indicate if this data is pre- or post-adjustment and use fewer than 400 stations when we all know 6 out of 7 of them are bum.)
EJ: “I am still seething over the USHCN-2 adjustments”
As am I but for DS’ experiment that’s not particularly relevant. However, maybe the results of his experiment could be of help in future adjustments? If so, I think the questions I’ve asked are pertinent to getting useful results (IMHO, of course 😉 )
Urban creep is a real issue. The first — and not least — of which is how to measure its impact. Obviously, the USHCN people think they have the answer. Considering the outlier effect of GISS reports, it, along with any Hansen adjustments, are likely incorrect. It would be interesting to hear their reasoning for some.
In some sense, DS’ work will not directly lead to proper adjustments. For example, it won’t help much on determining the impact of four parking lots placed 200 yards away but the results may indicate a course of action in investigating changes.
As am I but for DS’ experiment that’s not particularly relevant.
They say, effectively, that his experiment is invalid. It is NOT invalid. At least the slide show does.
(Those slide-show maps showing before-after are noticeably absent from adjustment page. As are the actual final figures.)
Evan, you’ve lost me. Am I missing some large elephant in the room or something?
‘They say … His experiment is invalid’ ??? They ??? Who are ‘They’?
‘slide-show maps’ ??? What “slide-show maps”?
Where did that come from and what does any of it have to do with Dave’s experiment?
I went back and read your previous post and may be you were still going on about it.
I’m afraid I tend to agree with the comment that the station siting isn’t particularly relevant if you are only looking for temperature differences. But, yes, urban creep will affect the readings. It’s a difficult problem and one NOT likely to be solved soon. I live at the top of a road that was two lanes when I moved here 30+ years ago and now it’s six lanes wide. In addition, this place was listed as being “in the country” — something hardly true anymore. I’d venture to say that there are few places on the U.S. East coast which WON’T have a similar problem.
In light of that, it’s doubtful that ANY adjustments to East Coast would be satisfactory — mainly, because I don’t see how anyone can state for certain that X development results in a specific Y bias. D. Smith’s experiment won’t help much in categorizing the effects of urban growth, either. For example, it won’t answer the question of what happens to a station that was once in the middle of farm country but is now less than a 2 sq. mi. island in a suburban development.
Still it might yield enough data for developing some insight into future adjustments, albeit, not explicitly. This would be especially true if it can be repeated in many places with more variation. That alone makes it worthwhile.
Probably the best approach for the USHCN is to carefully choose the sites it will use and make no adjustments at all. ‘Course, that opens a whole ‘nother can-o-worms 🙂
Scott, the third image is from Google Earth.
Gary, the heights of the two sensors are the same at 5 feet 6 inches (housing midpoint). I plan to test several heights to see how they affect the readings, out of curiosity. I’m sure that these height tests have been done many times by professionals but it’s enjoyable to replicate their results.
Regarding spikes, here are links to readings from today (hot off the press!)
Five-minute readings at each site are shown here
http://davidsmith1.files.wordpress.com/2008/05/0526082.jpg
Using the five-minute readings gives a 4F difference (95F versus 91F)between the max temperatures. If only the 30-minute readings are used (on the hour and half-hour), the spike is missed and only a 2F difference is recorded.
Same data, but showing just the differences between the two:
http://davidsmith1.files.wordpress.com/2008/05/0526083.jpg
An interesting (though small) image of today’s solar radiation two miles away is here
http://davidsmith1.files.wordpress.com/2008/05/0526081.jpg
The solar radiation drops when clouds pass overhead, as do the temperatures.
DAV, regarding the sensors, they are remarkably consistent. I’ll install the two sensors in the same housing and show the dead-on tracking. Kudos to the British firm which designed these little devices.
The ball field was picked because it provides a place far from asphalt and trees and is well-drained, and is within 300 feet (usually upwind) of the parking lot.
I say let’s have serious amateurs analyze climate for a time, at least they stick their hands out the window:
“2007 is likely to be warmer than 2006 and it may turn out to be the warmest year in the period of instrumental measurements. Increased warmth is likely this year because an el Nino is underway in the tropical Pacific Ocean and because of continuing increases in human-made greenhouse gases.” — James Hansen NASA/GISS, circa January 2007
“DAV, regarding the sensors, they are remarkably consistent”
Dave, I would expect them to be self-consistent with others of similar manufacture but that doesn’t mean they have linear response or match the readings of a known standard. Or did you mean they are consistent with some standard?
It may sound like a bother but showing the difference between these sensors and the one in the parking lot won’t yield a useful result if you don’t know the responses of all of them. One equation, two or more variables and all that.
Trying to figure out the response of the sensor at ‘A’ can be a problem. You need a sensor with known response in all locations. Is the parking lot sensor one of yours or is this a real station?
Even if you own all of the sensors, non-linearity can bollux your results — particularly if the resulting variation is comparable in size to the expected differences.
You also need to remove the housing as a variable. Placing all of the sensors in identical housing is a good way.
Sorry, by ‘A’ I meant the one in the parking lot. I see you’ve labeled that as ‘B’.
Nice work.
It would be nice to see a plot(s) of the temperature difference over time without averaging. The averaging is a little hard to get my head around in order to draw my own conclusions.
Also, I think you should include a “C-A” along with a “B-A” as some sort of experimental control.
Again, nice work.
‘They say … His experiment is invalid’ ???
They broadly imply that site quality doesn’t matter. It does. They say they properly adjust for it anyway. They don’t.
They ??? Who are ‘They’?
The NOAA
’slide-show maps’ ??? What “slide-show maps”?
http://wattsupwiththat.wordpress.com/2008/05/13/ushcn-version-2-prelims-expectations-and-tests/#comments
For the full slide show, the Rev posted the link in a REPLY in the comments section two or three posts from the bottom of the page.
I’d say David Smith has it right. His experiment roughly confirms Yilmaz et al (2008) which finds a great deal of difference in temp offset measured over grass Vs dirt Vs concrete/asphalt.
BTW, David, is how high is the sensor attached to the fence from the ground? Is it the regulation 1.5m? Are both A and B at the same height?
————————————————————-
To further the basic premise (and like so totally vindicate the Rev):
Yilmaz proves the offset bias. Kit proves the trend bias:
Ross R. McKitrick, Patrick J. Michaels , JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES, DECEMBER 2007, Quantifying the influence of anthropogenic surface processes and inhomogeneities on gridded global climate data
Abstract:
Local land surface modification and variations in data quality affect temperature trends in surface-measured data. Such effects are considered extraneous for the purpose of measuring climate change, and providers of climate data must develop adjustments to filter them out. If done correctly, temperature trends in climate data should be uncorrelated with socioeconomic variables that determine these extraneous factors. This hypothesis can be tested, which is the main aim of this paper. Using a new database for all available land-based grid cells around the world we test the null hypothesis that the spatial pattern of temperature trends in a widely used gridded climate data set is independent of socioeconomic determinants of surface processes and data inhomogeneities. The hypothesis is strongly rejected (P = 7.1 × 10−14), indicating that extraneous (nonclimatic) signals contaminate gridded climate data. The patterns of contamination are detectable in both rich and poor countries and are relatively stronger in countries where real income is growing. We apply a battery of model specification tests to rule out spurious correlations and endogeneity bias. We conclude that the data contamination likely leads to an overstatement of actual trends over land. Using the regression model to filter the extraneous, nonclimatic effects reduces the estimated 1980–2002 global average temperature trend over land by about half.
By half!
Just like I said back a-when! When Joe D’Aleo made that PDO/AMO correlation, and I commented that cutting the 1980-present increase by half (in honor of the Rev’s siting issues) would make the D’Aleo curve fit so much better! I just KNEW it had to be about half. It fit the picture too perfectly not to be. (They beat me up pretty good on Tamino for venturing that opinion.)
Now it’s Kit to the rescue! (Can I call ’em or what?)
————————————————————-
Great experiment.
Keep up the good work, David.
Interesting results… I’d like to see somebody set up about 10 or so of these temperature monitors (all calibrated of course) all around this field (or somewhere else with different landscaping), monitor the temperature and report the results. I’d really like to see it without wind too.
Any idea how the difference will be magnified or affected when the raw data is subjected to the various smoothing techniques and/or techniques to account for changes in sampling locations?
When I lived in southern Nevada, I noticed most people drove light colored, “desert friendly” vehicles. It doesn’t take much to also notice that people with lighter colored roofing had slightly cooler homes.
If everybody just covered their homes with aluminum foil in the summer (shiny side up, please), how much cooling would we notice? 🙂
Q&D calibration procedure.
1. buy a Styrofoam cooler.
2. Mount a low wattage incandescent light bulb in it.
3. Mount a small fan next to the light bulb.
4. Control the lamp wattage with a light dimmer
Take measurements over a range of temperatures.
This will let you know if the thermometers have different non-linearity characteristics. If you have a calibrated thermometer you can measure the actual non-linearity.
One thing to watch for. Electrical noise. Dimmers are notorious. A filter can help. Possibly a bit of tin foil around the USB connector (not touching the pins) can help.
More sophisticated tests can be done with a similar set up: DC supples, temperature controller, etc.
David,
Your guest entry is interesting and your work inspirational as always. If I may ask a question on method: How did you isolate sensor B temperature data for periods of exclusively north-blowing winds or south-blowing winds?
Thanks,
Bill P
“Q&D calibration procedure”
Not a bad procedure although I would use a solid state warmer. They’re often found in battery operated six-pack coolers. They’re adjusted the same way, i.e. by changing the duty cycle, but not as noisy.
DS still needs to characterize the parking lot sensor (if he hasn’t done so already) or the results aren’t very meaningful because:
Let TP = temperature indicated by the parking lot sensor.
TP = Tp + Ep + EP, where
Tp=true temp (ambient) at the sensor (if the parking lot wasn’t there)
Ep=sensor error
EP=error introduced by the lot and surrounding terrain
Assuming DS knows the true temp of his own sensor (let’s say at A and called TA), to find EP, he needs to calculate:
EP’ = TP-TA = Tp+Ep+EP – TA
The built-in assumption is that Tp and Ta (the true ambient) should be the same but this is likely true.
See the problem? There are two unknowns (Ep and EP) in that step so unless he has really knows the sensor characteristics at the parking lot he’s mostly wasting his time. What he’s finding is (Ep+EP), which is what he set out to do.
Here’s a question I always wanted to ask someone and this crowd seems like the right one to answer it:
Would it be possible to select, out of all the temperature stations in the world, 100 or so that are undeniably way out in the country removed from any possible urban creep effect, and which have been in the same place for decades, and plot the historical average temperature from these stations? Basically create a “pristine” average temperature record?
Any comments appreciated.
Sean
REPLY: I’m working on that, see http://www.surfacestations.org – Anthony
They also sell an “Ethernet and Internet Ready Thermometer with built in web server” that has accuracy to .1 that uses separate thermistors as probes, but it’s a bit pricy.
These devices have a 1 F internal resolution and a +/-2 F accuracy.
RE: Mark O (13:56:56) :
Gage R & R is treated, oddly, as a sort of heresy by most orthodox so called “Climate Scientists.”
Hi Anthony, thanks for the link. I have visited your site before; it’s a great project.
I couldn’t actually find any mention on the site the idea of a global clean temperature record, but I suppose after you’ve identified enough sites it will be easy to construct.
Any idea how many (true) Class 1 sites you would need for the result to be statistically significant? Maybe you have enough already?
Sean
It’s not the number, it’s the geographical distribution.
CRN is supposed to address this. Here’s hoping.
David,
Re: satelite view (wider field of view)
I am curious to see what is around your chosen site, besides lots of barbed wire.
Best,
Scott
[…] Up With That they’ve got some preliminary results from a study of what they’re calling “The Parking Lot Effect.” Follow the links for the details, but basically a paved surface like a parking lot can bias […]
[…] http://wattsupwiththat.wordpress.com/2008/05/26/the-parking-lot-effect-measurement-bias-in-locations… […]