Yes, it’s back. Dr. Roger Pielke Sr. and I have started looking at the GHCN to document the siting quality of surface stations that measure climate in the Global Historical Climatological Network (GHCN), the darling of NCDC’s Dr. Thomas Peterson who has recently added another level of processing to the mix.
I’ll be offering some instructions soon on how to participate. In the meantime, feast your eyes on this gem, the GHCN station in Tampico, Mexico, #76548. What is it with air conditioners and weather stations anyway?
NCDC has pretty much non-existent metadata on it. Even the start date is wrong.
The lat/lon is so coarse that it pretty much is useless, as Steve Mosher demonstrates, it makes Hansen’s nightlights UHI adjustment pretty way off-center. But hey, “close enough for government work“.
Hopefully, they’ll bother to put in a “GHCN” flag like they did after we embarrassed them into doing in the original surfacestations project for USHCN. As a results of the efforts of volunteers, we actually had better metadata than NCDC did, and they couldn’t have that, so they fixed their own problems by making the lat/lon more accurate and putting USHCN flags into the database. It is a far cry from their first response, which was to block the database for our use. I won that argument hands down.
Here are some other examples in Mexico:
Veracruz, 76692. I wonder if they have a cat on that tin roof at night?
Salina Cruz, #76833:
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
stumpy says:
August 29, 2011 at 12:32 pm (Edit)
A list of all GHCN stations by country would be useful, I know quite a few of the ones in New Zealand well, but I am not sure which ones GHCN uses. I already have pics of some of them and I believe there is a book available in NZ with station history for most of them.
####
It depends on the dataset you are looking at but you can get the country several ways.
In GHCN the country is encoded in the ID
In GCHN daily the country is field in the metadata.
You can also send the coordinates to any geocoder and get back the country. Use geonames.org and its relatively easy if you know how to program. If you dont know how to program, you’re SOL
george
“Temperature; that shouldn’t be too difficult since according to Dr James Hansen, it is close enough if you are within 1200 km of where you want to know the Temperature”
you would actually be surprised how consistent temperature anomalies are within a 1200km radius. Like you I thought this was suspect, Looking at the data I was convinced. One meaning of the word climate is this: we know the winter climate of florida is different than the winter climate of arizona. location matters. How much does it matter? we can measure that by looking at the correlation length. turns out, for some parts of the globe at some times of the year, 1200km is a good number. During other seasons and and different latitudes it can be as high as 1500km. Some places more around 800km. It’s just data you should look at it
What I found interesting was the assertion that siting issues didn’t “corrupt” the warming trend. I guess it goes to show the old saying “they don’t make ’em like they used to” is true after all.
All those medieval air conditioners, jet aircraft exhausts etc running since the pre-industrial age obviously wouldn’t affect the trend.
I see a picture of a box and some anenometers. But what’s the basis for believing that it is GHCN station #76548?
Rosco says:
August 29, 2011 at 2:40 pm (Edit)
What I found interesting was the assertion that siting issues didn’t “corrupt” the warming trend. I guess it goes to show the old saying “they don’t make ‘em like they used to” is true after all.
###############
It’s not that surprising of a result.
Let’s start with the basics. In the Field tests of the development of the CRN scale ( 1-5) the researchers found the following:
1. the micro site effect was modulated ( increased or decreased ) by factors such as wind, rain, and clouds.
2. the primary effect was to increase the RANGE of highs and lows and not the MEAN.
CRN 3, for example, refers to the fact that the station may see highs on certain days
that are 3C higher and lows that are 3C lower. It’s a measure of variablity, not BIAS
3. The bias was roughly .1C warming for all classes of stations.
basically you have a sporatic effect that is modulated by wind and clouds and rain. Its the same with UHI. For example in some locations a surface wind of 2m/sec will mitigate UHI, in other locations 7m/sec erases all the UHI. Is there UHI and microsite? sure. That’s not the question. The question is how large? how often? and which direction is the bias?. The answer to date is as follows
Not large enough and not often enough to create a significant detectable difference in global trends. no cookie, yet
They closed the U of A parking lot station! Damn!!! It’s my favourite one.
Brent in Calgary
I see a picture of a box and some anenometers. But what’s the basis for believing that it is GHCN station #76548?
#####
Good catch Nick.
Looks like there is some confusion about that site. The system that Anthony is using (MMS) looks to be pulling up WMO data. Thats the location of the WMO station. which cross checks with publication 9.
However, that station does not get used in any of the official datasets.
Its not in GHCN Daily. and the Tampico station that does get used in GHCN Monthly
is at a different location.
41476548000 22.2200 -97.8500 9.0 TAMPICO, TAMP 32U 212FLxxCO 5x-9COASTAL EDGES C
So, 1. we dont know that the picture is of the location and 2 we know that the station at the location given by MMS doesnt make it into the global average and 3 the station in Tampico is classified as Urban.
“I’ll be offering some instructions soon on how to participate. In the meantime, feast your eyes on this gem, the GHCN station in Tampico, Mexico, #76548. What is it with air conditioners and weather stations anyway?”
As the MMS system shows NCDC have no temperature data for this station.
In short, the bad station contributes no data. That would be a good thing.
It also looks like the station only has data since 2007 to present, which means that it would fail the 10-year minimum requirement for inclusion in GHCN (as well as the lack of any common baseline period years in a CAM approach like NCDC uses).
REPLY: No, that’s the add date at NCDC, if you’ll check around, you’ll see many GHCN stations that started in 2007, prior to that they had no record of these at all in MMS – Anthony
‘Steven Mosher says’
Points well taken but it still comes down to site maintenance, calibration and changes that have occurred around the site since day1 etc…. just because you can get an average over 1200km says nothing about these points. The surface data is suspect to say the least.
tom s says:
August 29, 2011 at 4:14 pm (Edit)
‘Steven Mosher says’
Points well taken but it still comes down to site maintenance, calibration and changes that have occurred around the site since day1 etc…. just because you can get an average over 1200km says nothing about these points. The surface data is suspect to say the least.
#######
For example, recently to test some new algorithms ( created by Nick stokes, RomanM, and Tamino) that have a sounder statistical basis than GISS or CRU, I looked at a 100 stations in
texas. Just testing these algorithms. Any way, when I was done testing I asked myself, I wonder how well correlated they are? I mean if we assume that there are calibration problems all over the place and bias problems and all sorts of data horrors those horrors should show up in time series
correlations. right? No odd balls. 98% correlations across the board. no odd balls. But texas is not the world, its just a big area I selected randomly to test the algorithms.
let me repeat this yet again. To bias the anomalies you have to have widespread biases with a consistent directionality. An odd ball here or there is washed out in the global total. we dont see widespread oddballs. The most interesting unanswered question centers around the issue of categorizing stations into urban and rural classes. That was the question 4 years ago, its still the question today. Answering that question depends upon.
1. having accurate station location data
2. having physically relevant and objective proxies for urbanity
3. categorizing stations while minimizing false identification and maximizing global coverage.
Good timing.
I’m off to Morioka (in Tohoku) tomorrow. I’ve located the station 21047584001 MORIOKA near where I’m staying, and printed off the map. Photos and site description to follow. I’ll take my tape measure.
Cheers
Hector Pascal says:
August 29, 2011 at 8:44 pm
If you can geocode your photos even better.
So, Anthony, is this a worldwide surfacestations project that you’re proposing? I note from the list of GHCN stations that there are 6 in the Kamloops area which I can get out and photograph assuming the lat/long coordinates are sufficiently precise. I know of one station at the university (TRU) which seems to be poorly sited as it’s bordered on the west by an asphalt roadway and parking lot. Not sure if it predated the university expansion.
If I’m using my GPS (which is a USB module) I’ll also be able to run some temperature transects in Kamloops as I’m driving around. That is, if I can find some more batteries for the USB temperature monitors as mine have finally all died.
REPLY: Yes another social networking project, like before, but this one has greater challenges. Details to follow in a few days – Anthony
Greg, Spokane WA says:
August 29, 2011 at 10:26 am
“Any thoughts on how much that air conditioner or tin roof might skew measured temps?”
Tin roofs of the variety shown get hot as Hell and hold it for a long time. For comparison, as long ago as the Sixties, we had to install insulation underneath tin roofs on “chicken houses” even though the peak of the roof was 14 to 16 feet above the floor of the chicken house and, in warm weather, the chicken house was entirely open to air front, back, and sides. The area was central Georgia USA. A “chicken house” of that era was about 210 feet long and 40 feet wide.
Of course, the Warmista will reply that the tin roof does not matter because absolute temperature is irrelevant and only the trend matters. If you follow this Warmista reasoning, no location of any temperature station can have any adverse impact on temperature trends from that station. Warmista are so lacking in the ability to reason. If they take their principle seriously, the only possible conclusion is that all temperature readings and temperature trends are worthless. Why? For the obvious reason, they are not responsive to what is happening in the environment.
Maybe I should add a bit more explanation. Take the clear case of air conditioners on private buildings. Warmista assume that the air conditioner will be on at all and only the same times every day or they assume that it will be on at random times, in the sense of randomness used in statistics. Now, tell me, what justifies such assumptions? Obviously, absolutely nothing. What goes against such assumptions is bedrock knowledge of human behavior. Human behavior is erratic or chaotic but not random.
Steven Mosher says:
August 29, 2011 at 2:18 pm
” you would actually be surprised how consistent temperature anomalies are within a 1200km radius.”
Steven Mosher says:
August 29, 2011 at 4:32 pm
“For example, recently to test some new algorithms ( created by Nick stokes, RomanM, and Tamino) that have a sounder statistical basis than GISS or CRU, I looked at a 100 stations in
texas. Just testing these algorithms. Any way, when I was done testing I asked myself, I wonder how well correlated they are? I mean if we assume that there are calibration problems all over the place and bias problems and all sorts of data horrors those horrors should show up in time series
correlations. right? No odd balls. 98% correlations across the board. no odd balls. ”
=================================================================
98% correlation across the board between stations is achievable only if one fails to remove the overall mean of the yearly average time series, or if the annual cycle–which dwarfs climatic variations by orders of magnitude –is not removed from monthly averrage time series. Done correctly, inter-station correlations are very much lower than that. Taking the yearly data at Coriscana TX as an example, one finds R^2 estimates of 0.61 with Corsbyton (473km), 0.22 with Eagle Pass536km) and 0.35 with Alpine (716km). That’s a far cry from the 1200km radius that is assumed to be the effective “correlation length.”
I found station 21047584001 MORIOKA.
Nada. I couldn’t believe it. About a 10mx10m block on the side of a hill in a heavily suburbanised part of town, completely surrounded by 2-3 story buildings. I thought “I must have made a mistake”. Home, I’ve checked the data, and I had found the location correctly.
Nothing there. It’s been moved. That site would have failed meteorology 101. Sorry folks, my Japanese isn’t good enough to go knocking on doors.
There is a bottom line to all this, Mr. Mosher. Two bottom lines, actually. One is that your attitude seems to be that it would not matter if a weather station were mounted on the hood of a 1956 Buick Roadmaster Anniversary Edition. In other words, you show total disdain for the facts on the ground. And you have a meta-attitude of triumphalism, happily explaining that the statistics can overcome everything.
The second major bottom line is that you have no clear idea of “climate” as opposed to “weather,” a problem that haunts all Warmista. If you cannot independently specify what you mean by climate then we remain clueless about the meaning of your statistical claims. But I do not believe that you can do this because “independently specify” means what it has always meant in science, namely, specify in accordance with scientific method through hypotheses that have shown reasonable confirmation.
I’m sorry Steven Mosher, we must have you out of context here. I am very sorry.
I assume what you mean to say is “Wow I didn’t realize how poorly those stations are sited. We really need to put some money into infrastructure. I still think the data are good enough for the purpose but better data would be important.”
Right? That’s what you mean isn’t it? You can’t possibly be saying that these stations represent no significant accuracy problems statistically speaking and that you see no reason to correct these discrepencies. You really aren’t saying the latter are you?
Is the Scientific Method really just a loose set of guidelines that scientists don’t need anymore?
Your slip is showing.