An interesting post by regular CA commenter David Smith at Anthony Watts on his results quantifying local microsite effects on a thermometer placed near a parking lot. Interesting to compare this with efforts of Jones, Parker, Peterson etc. to “prove” that site effects don’t matter.
-
Tip Jar
(The Tip Jar is working again, via a temporary location) -
Pages
-
Categories
-
Articles
-
Blogroll
- Accuweather Blogs
- Andrew Revkin
- Anthony Watts
- Bishop Hill
- Bob Tisdale
- Dan Hughes
- David Stockwell
- Icecap
- Idsos
- James Annan
- Jeff Id
- Josh Halpern
- Judith Curry
- Keith Kloor
- Klimazweibel
- Lubos Motl
- Lucia's Blackboard
- Matt Briggs
- NASA GISS
- Nature Blogs
- RealClimate
- Roger Pielke Jr
- Roger Pielke Sr
- Roman M
- Science of Doom
- Tamino
- Warwick Hughes
- Watts Up With That
- William Connolley
- WordPress.com
- World Climate Report
-
Favorite posts
-
Links
-
Weblogs and resources
-
Archives
Climate Audit 
32 Comments
Anyone who thinks site violations “don’t matter” needs to read this:
Yilmaz et al (2008)
Click to access ATM002100202.pdf
Help sought. With he increased use of vehicle Satnav and the near universal presence of car thermometers, I have been mentally formulating a plan to do a temp survey of my home city of Melbourne (Australia, not Florida).
The broad plan is to capture a sponsor with both communication and computing facilities. To interest a sponsor, one needs to show possible benefit from partaking.
Has anyone knowledge of this alrady done elsewhere so the trials and pitfalls can be avoided? There was a CA post a few weeks ago where home weather stations were used, but there are not so many here.
I’m thinking of approaching the State roads and motorists’ association, but first I need more info on the response of car thermometers. Is it best to read them before starting a cold engine, or after half an hour of driving, in shade or sunlight, over asphalt or grass etc? The aim is to produce a series of contour maps for different days with the expectation that UHI will stick out like a bullseye with numbers on it.
Suggestions?
I don’t expect to see such variation as in Turkey #1 above. That is astounding and it is an airport. Does anyone know of studies of temp changes with distance from runways and taxi ways? Australia’s high quality system uses quite a few airport sites.
I don’t know if car thermometers would be a good idea, unless you can quantify any effects the car might add. Where the sensor is on the car is important, obviously.
Last winter when I got my first car with an outside air thermometer, I was very excited about it. I live dead in the centre of Amsterdam, and work quite some distance outside the city. I noticed that in the morning, when you’d expect temperatures to rise, it would generally get colder when I left the city. In the evening, when you’d expect temperatures to drop, it would generally get warmer as I entered the city. This would be confirmed especially in the morning when the temps were around freezing, by the fact that inside the city parked cars would generally not have frost on the windows, while cars outside the city would have.
Also on my holiday in Sweden, we’d spend the nights in a cabin well outside a city, and as we’d drive towards the (relatively small) city, it would always gets warmer, no matter if it were morning, evening or anything in between. And it’d always get colder as we did the reverse trip. Terrain changes may have had a higher impact than UHI, though, I’m not sure.
Since spring, I can’t really detect such a clear signal. Sometimes temps rise, sometimes they don’t. When I’m stuck in a traffic jam, it immediately affects the reading – I’ve noticed the difference between standing still and moving at highway speed can be as much as 1,5 degrees.
Maybe I should make regular notes instead of just looking at it every now and again, to get a better idea of what causes what. Even when parked there are great issues, but when moving with different speeds through different terrains at different times these issues become huge. It’s fun to look at, but I doubt you’ll get very reliable results.
Re car thermometers
I realise the limitations expressed (as I have been thoughtful about the thermometer response of my car, a fuel efficient 3.8 litre V6 engine with a supercharger), but given a huge public response most deviations will end up as noise and might be not so important. There’s no shortage of statisticians on CA to make error bars. Most of them even know how to do it properly. I am hopeful that the signal will be quite a lot larger than the noise. Help is asked for as I have no experience below about 2 deg C (say 35F) – do cars for quite cold areas have thermometers in different places to avoid snow lodging, that sort of thing. My main worry is saboteurs phoning in with false readings during the event. Which is a variation on modellers “adjusting” their false readings after the event.
Here in the UK weather forecasters accept that there is an urban heat island effect. On more than one occasion this year I heard a forecaster say something like:
That is a massive 4C temperature difference between rural and urban which is why I remembered it. I’ll bet if you asked the same people, who produced the forecast, about UHI they would claim its effect was minimal.
My own personal experience is that generally there is a 1 to 2C difference between where I live (on the outskirts of a city) and work (about 4 miles from the center).
A college blogger did this type of survey a while ago. It provoked quite a storm. I will try to google it.
Geoff http://www.climate-skeptic.com/2008/02/measureing-the.html
Terry#6:
Occasionally you can see urban heat island effect on the temperature contour maps of the London area for the BBC (spits) weather forecast. London sticks out like a big warm oblong.
So UHI is clearly built into the meteorological models.
Now, why would it not be in the climate models? Does the right hand not know what the left hand is doing?
Geoff: all you need is enough participants and the law of large numbers will make those error bars shrink towards zero 😉
In the imortal words of George Kastanza: “It’s not a lie if you believe it.”
Here’s a plot of the parking lot effect:
The chart should be self-explanatory, but here’s an explanation anyway. The grey line is the raw HadCRUT3 global monthly anomaly, and the blue line is HadCRUT3 global monthly anomaly smoothed. The remaining lines attempt to break out various components of the “global” warming represented in the global temperature “signal.” The smoothed trends for the tropics (in HadCRUT3, this is 30S-30N) and HadSST2 for the Northern Hemisphere generally follow the “global” trend. The exceptions are HadCRUT3 and CRUTEM3 trends for the “Northern Extra Tropics,” i.e. north of 30N. The trend for HadCRUT3 Northern Extra Tropics (green) will contain both land and sea temperatures. CRUTEM3 Northern Extra Tropics (violet) is land only.
It seems to me that “climate science” has some explaining to do. Is land north of 30N becoming one huge parking lot? Looks a little that way, doesn’t it?
Regarding temperature sensors integrated into vehicles: I had a Chrysler Town & Country with a sensor that was extremely responsive and accurate and did not seem much affected by engine temperatures – always moving forward when engine temperatures were high, sensor appeared to measure air temperature rather than IR emissions from the engine. Current vehicle is a Nissan with a fairly pathetic sensor. May or may not be accurate – the problem is the response time is on the order of 15 minutes.
Come on, UHI doesn’t exist and there is no work on the subject that we know of.
There’s no way any of this could have an affect upon climate.
– Size and shape of cities—aerodynamically, cities have a very different shape than rural areas. Tall buildings act as obstacles and reduce wind speeds.
– Urban deserts—cities can be thought of as virtual deserts with almost no vegetation and materials that are almost completely impermeable to rain. This combination leads to a lack of evapotranspiration which increases sensible heat.
– Urban canyons—the tall canyons formed by city buildings trap radiant energy in their walls. Comparisons of this “canyon effect” in European and North American cities suggest that areas with denser and taller buildings will more rapidly develop heat islands.
– Humidity effects—although there is little difference in the amount of water that cities and countrysides retain in their atmospheres (absolute humidity), the higher urban temperatures effectively lowers the relative humidity (since warm air can hold more water than cold air).
– Urban haze—The haze of air pollution that hangs over many cities can act as a miniature greenhouse layer, preventing outgoing thermal radiation (heat) from escaping from urban areas.
– Anthropogenic heat—The release of heat from the burning of fossil fuels and also raise urban temperatures. The human-induced energy release over Manhattan Island is approximately 400% the insolation on a winter day.
NASA Earth Observatory – Urban Rain
The Urban Heat Island Effect And Its Influence On Summer Precipitation In The City And Surrounding Area
Here Comes Urban Heat
Urban Climatology and Air Quality
Keeping New York City “Cool” is the Job of NASA’s “Heat Seakers”
Weather Matters; George Taylor, Oregon State Climatologist
The Big Picture Lady
U.S. EPA
TRMM
PDF 1
PDF 2
PDF 3
PDF 4
PDF 5
PDF 6
PDF 7
First link didn’t work.
http://earthobservatory.nasa.gov/Study/UrbanRain/
Earth Observatory – Urban Rain (2006)
I recognize it’s a way of expressing exasperation, but don’t you think your formulation as “prove” might weaken your point? (Assuming that it’s not a direct quote from a journal article title “In Whiche We Prove Beyonde Alle Doubte That Site Effects Do Not Matter” or some such goofy thing, in which case I’ll switch to “don’t you think it would be more persuasive if you backed up the goofy-sounding direct quote with a hyperlink to the goofy original?”)
I nominate “Interesting to compare this with efforts of Jones, Parker, Peterson etc. to demonstrate [or ‘show’] that site effects are negligible”.
My friendly GMC dealer informed me that the outside air temperature for my Suburban is measured via the Intake Air Temperature (IAT) sensor. This sensor is located in the air intake manifold. On some cars it is combined with the Mass air Flow sensor. The air is drawn in to the engine via the intake manifold on the front of the car behind the grill. This ensures that the reading represents the ambient air temperature and is not affected by the sun, engine heat, snow buildup, rain, etc. It is vital for proper air fuel mixture. It also regulates how the A/C mixes outside air with the cabin air. The reading that you see on your outside temperature gauge gets its data from the IAT via the car computer.
Has anyone out there ever done any sleuthing work on the Heathrow (London ) temperature gauges. As far as I can see from the coordinates and Google earth it is very close to the runways in a very built up area. This is regularly cited as the warmest place in the UK.
Tony B
Here’s a time series which I find interesting:
This shows the “average daily temperature” of the baseball field calculated in two different ways. The blue is an average of the 48 readings taken 30 minutes apart. The mauve is the average of minimum and maximum, the method used by climatologists. I consider the average of the 48 readings to better-represent the average daily temperature than the average of min and max (though I suppose that is arguable).
The two methods seem to agree closely at the lower temperatures but diverge at the higher temperatures.
This divergence may be due to changes in the types of air masses present, or maybe due to something entirely different, or may be it’s a data ghost. I don’t know which.
The interesting question this divergence raises is whether the relationship between the two averaging methods (2-point and 48-point) is something other than 1:1 over some temperature ranges. What if, in the tropics, the “true” (48-point) average rises 0.10C but the “standard” (2-point) method reports a 0.15C rise? Would the “standard” climatological method be exaggerating the “true” tropical temperature increase?
This apparent divergence, if it is real, also raises a potential caution flag or two on interregional and interseasonal temperature comparisons.
I have no opinion one way or another, as this plot covers only sixteen days in May 2008. But it’s something to ponder.
You make a great point, David. Reminds me of my thought that if an area spent 1 hour at the min but 6 hours at the max, where’s the beef?
Only put my dog in those parking lots, watch where he chooses to reside and not reside, and you know a lot about temperatures in that microclima.
The Barkihg Lot Effect.
#14 (Sam) says: “Come on, UHI doesn’t exist and there is no work on the subject that we know of.” and follows with excellent examples and links which imply otherwise. David’s study is interesting and probably quite accurate. Ared’s car study is interesting too and probably roughly accurate. But David is measuring a local site effect and Ared is roughly measuring a regional UHI effect. The site effects cause the UHI effects. Attempts to compensate for UHI (e.g. the GISS adjustments) are hopelessly boggled by site effects which may reverse the UHI effects. In any case the proxies for UHI are often inaccurate (e.g. out-of-date population figures).
I think the climate community could use the studies like what Geoff is proposing as well as models with microscale grids. UHI certainly has an effect on real world climate as urbanized and suburbanized regions grow larger. But the more important effect of UHI is mixed with the site bias effect, namely the statistical and ultimately political effect of biased reporting.
When it comes to using only temperature differentials, it really doesn’t matter where the sensors are placed provided that the site itself doesn’t contribute to variations in the deltas (e.g., growing urban development). Ared’s experiment results show this: he’s reporting deltas. The temperature errors caused by placement cancel out. The average temperature differential of a parking lot thermometer should be constant in an unchanging environment. IOW: it’s self-calibrating. So what if it’s X degrees higher (or lowere) than the surrounding area? Thermometers in both locations should show the same delta.
It’s a completely different matter when you try to use specific temperature readings vs. differentials (e.g. today’s temperature is xxx).
David Smith’s experiment, while otherwise well conducted, has a disconcerting potential flaw that would render the results unusable. It’s not clear that he took the necessary steps to avoid the flaw.
DAV, not sure what your point is. You say
But that’s exactly the point. We don’t know whether site X “contribute to variation in the deltas”. However, the assumption by many AGW folks (that mostly it doesn’t affect the deltas) seems to me an unproven hypothesis rather than an observation.
In the case of the parking lot, you are correct that once the parking lot is established, you may not see much difference in the deltas. The problem, of course, is the jump that occurred when the parking lot was built, which definitely affects the deltas.
w.
Re: #23
I have not chimmed in here yet as I thought David was getting all the “help” he needed (or could stand). I have not a clue to what DAV is eluding to here. I think David is collecting some temperature data under well measured and observed conditions and not necessarily to provide a basis for a peer-reviewed paper. I think the analysis of these data will provide some interesting insights into what could affect temperature readings on a micro-site scale. I particularly liked the wind influenced temperature and now determining whether that is a micro-site effect.
One can look at the magnitude of micro-site effects prevailing at a given time and then make assumptions on whether that environment occurred in one momentary step (and was or was not detected by looking for break points in the time series) or whether it could have evolved over longer times and making detection of the change more difficult to detect (and adjust for). Anyway it all starts with good and well-controlled tests results.
DAV says:
The question would be if the delta between 5 different calibrated and unbiased thermometers a few feet to a few hundred feet apart in an unchanging enviroment would all be the same delta.
In order to start on that, you’d have to have 5 unchanging environments about the same distance away from each other. Probably only the best that could be hoped for is 5 that change at the same rate.
So it seems this would be difficult to do as a controlled experiment without a lot of attention to detail.
The point being, the delta around a swampy grass area with changing wind patterns would not be the same as one near a fairly static parking lot, one wouldn’t think. Just the materials!
Then you have the problem I suggested; if a day is at the minimum temp for 1 hour and the maximum temp for 6 hours (or vice versa) does that really tell you anything to get the mean of min/max?
Re # 8 MarkR
Thank you, that is exactly the type of feedback I was seeking. You gave me a feel for the amount of variation I can expect and that influences the design of the experiment.
To interest a sponsor, I have to show that a better result than nil or uncertain can be (and has been) obtained. Much appreciated.
@24, Willis:
I outlined it in the comments at WattsUpWithThat. Basically, Dave had published the difference between two sensors (one in the parking lot and one on the ball field). Presumably he is claiming to be showing the effect of the parking lot but the parking lot sensor has at least TWO things that could affect its readings. This means he has two variables (at least) with only one equation. That makes the results useless IMHO.
He may have been able to correct for this but AFAIK he hasn’t said so one way or another (I haven’t checked recently). I suggested repeating the experiment using his own sensor.
Yes, that was my point. I thought it was what the experiment was all about. It has the potential to allow development of better adjustments for changing local environment.
#26 Sam,
The time period should also be considered. I’d think you’d have to take averages spaced by a year to get really meaningful results.
#24, Willis
And we still don’t although there’s data there for a quick look.
The experiment will take at least a year IMO and should contain data from at least two places. Sam U suggests 5 and I tend to agree but that may be expensive. It also needs to be repeated to remove experimental error.
Can I bring in the question of how few weather stations/temperatures nodes there are in reality, which must affect the records drastically, quite apart from the accuracy of the data being recorded. From what other posts have said the number of weather stations is now around one third of what it was in the 70’s. That means we are missing an awful lot of weather (perhaps more than we are capturing?)
As an illustration, I have just returned from my drive to work (southern England). After a mile or so of fairly good dry weather the skies suddenly darkened and the rain literally fell like stair rods, to the extent I had to pull over for ten minutes. I continued when the rain had eased off a little. Coming round the edge of the city in this terrible weather there was literally a line in the road and on the other side it was dry and bright with people mowing the lawn. Two miles further on the line of rain reappeared and the rain was of biblical proportions with very severe road flooding. I got home and it was only raining lightly having been dry all day. The rain (not a thunderstorm) straddled both sides of the estuary I drive round and extended for perhaps 10 miles square. I happen to know there are no weather stations in the area so it would go unrecorded. The temp in the rain was 11C, the other side on the dry line it was 19.5C according to the car thermometer.
I don’t know whether satellites can genuinely claim to capture temperature records but it was certain that the area I travelled in would be recorded by the weather station ten miles outside the rain area as being dry and warm. I don’t know if anyone has ever done any calculations on the number of weather stations we need globally to accurately record the huge local varaitions in weather, but I suspect it is many times what we have got!
Tony B
Hi, DAV, and thanks for the feedback.
You are correct that there are multiple variables affecting the parking lot thermometer (and the field thermometer for that matter). Cloud cover patterns, sun angle and duration, humidity, instrumentation variability, windspeed, wind direction, day of week (parking patterns), pavement albedo, moisture content of nearby soil, current rainfall, grass height, ground cover type, nearby vegetation and so forth are variables.
The temperature profiles of the seven days in May do not define how parking lot proximity affects temperature. But, they do hint at the complexity of the situation and that antropogenic effects like the parking lot matter.
As you indicated, it will take multiple sensors and extended runs, stretching across seasons, to more-fully characterize anthropogenic effects like parking lots.
These parking lot results are “practice”, as I mentioned at Anthony’s site, wherein I work on techniques and I get a feel for the equipment and data. The reasons I illustrate some of the results are to draw helpful comments from readers and (I hope) to encourage others in other regions to make their own microsite experiments.
And, these practice runs can create intriguing patterns, like the one mentioned in #19.