Apr 23 2008
Arizona Senora Desert Museum aka CRN Tucson 11W
Yesterday I did a post called Revisiting Tucson USHCN: Regional Correction Factor. I thought it would be my last post on the UofA surface station in a while, but I guess not. The US Climate Reference Network (CRN). This is described as “a network of climate stations… Its primary goal is to provide future long-term homogeneous observations of temperature and precipitation that can be coupled to long-term historical observations for the detection and attribution of present and future climate change.”
Luckily for me, steven mosher pointed out that there is a CRN station really close to Tucson at the Arizona Senora Desert Museum. He said:
last time I looked at the UofA data Atmoz, I recall [the Arizona] senora [Desert Museum] being an intresting comparsion. I might be misremembering, but did you have a look at that?
So I finally got around to downloading the monthly data for this CRN site. This wasn’t as easy as it sounds since they have not yet provided a nice interface to do this, or provide all the data in one file like the USHCN.
The first full month of data for the Tucson CRN site is in October of 2002. I haven’t updated my USHCN data is a while, so the last month of data is in May of 2007. But that’s still 4.5 years of data, which allows us to compare the two.
|
Here we can see clearly that yearly cycle in both of the temperature time series. The temperature at UofA is slightly higher than at the Desert Museum. Note I used the mean monthly value reported by the CRN and not the (max+min)/2 value, so there may be slight differences in monthly values. However, the correlation between the two time series is extremely high.
Looking at the monthly anomalies will allow us to determine if there are any siting issues that appear in the UofA record for this time period.
|
Again, the correlation coefficient is extremely high. The anomaly time series follow each other almost perfectly. I think we can reasonably conclude that whatever microsite issues may be present at the Tucson UofA surface station, that for the time period used in this study that they are extremely small.
(No Ratings Yet)
Loading ...Related Posts:
18 Responses to “Arizona Senora Desert Museum aka CRN Tucson 11W”
To reduce spam, comments are automatically closed 30 days after the last comment. If you would like to comment on any closed thread, please use the contact form at the top of this page.
Yes, now one thing I have been trying to get across to people is that Microsite bias is real, But it will only impact your trend estimate if you take a trend over the period in which the bias occurs.
In short if the bias at the University happens at say 1998, then
after 1998 I would expect its trend to track the trend of surrounding stations. The bias is just an offset. If I compute trend for the bias station starting at a date before the offset ( say 1980)
then the offset will hit the trend estimate of course. and if we want
to compute long trends accurately they we want to fix as many of these little biases as practicable
All of which means researchers need good documentation of stations history. Even starting today with the kind of documentation that AW is doing.
Run Senora against the airport, I think you’ll find the same thing.
Oh ATMOZ, St.MAC has a script somewhere for scraping CRN data, It’s a bitch doing it by hand. Hopefully Karl et al will give FTP access.
[Reply: It would have been easy to write a script to do it, but their robots.txt file excludes that directory. And I would have needed to parse their html.]
You did by hand?, I thought I was the only crazy person. Probably
why I remember senora after all this time. Glad to see you over at
Lucia on the satillite thing. She is wicked smart kiddo, and a nice person.
Anyways I reminded Anthony to ask for better access to the CRN data. Hope springs enternal for data junkies.
mosh, you’ll be beating your head against that wall for a while longer, I’m afraid. Those folks have themselves completely convinced that the data from any station with a demonstrated microsite bias must be completely junk. This is the result of blindly adapting the CRN siting standards to grading USHCN data quality.
Just out of curiosity, does the CRN site give you the temp data already merged or is it possible to see the readings for the individual sensors? Also, I notice that they have an IR sensor for ground temps. Is that used for some sort of adjustment for the variation in ground cover (which per the station photos is considerable)?
While I’m at NCDC, they have promised to setup an FTP data acess for CRN data, so that you don’t have to scrape or hand enter CRN data.
I’ll make that available as soon as they provide it.
[Reply: Thank you!
]
I know this is a very short time frame (climatically speaking). Could you show the differences between the anomalies or the trends of each? Eyeballing it, it looks like a slight difference in trends, but that might just be the calibration on my eyes.
Steve Bloom, yes i think you can get traces from all three temp sensors
( memory again, so you should double check)
here.
http://www.ncdc.noaa.gov/crn/hourly
there is another page of data somewhere
A script to scrape individual CRN stations is at http://www.climateaudit.org/scripts/station/read.uscrn.txt . I haven’t reviewed it recently as sometimes things change, but it has some handy devices for scraping the html pages. I can see no reason why a robots.txt statement should be construed as a barrier to a researcher downloading data. I had a short dispute with NASA GISS about this, but they conceded and I’d be astonished if NOAA would construe their robots.txt as applying to a researcher.
Steve Bloom, My head and walls? no contest.
take that as you will.
BarryW I think the first thing you want to look for is obvious
discontinuties: jumps. jolts, jerks.
Then I would check for creeping TMin, minumum temps rising.
Then I would check for creeping Tmax.
By the time you check the trend in (TMAX+TMIN)/2 the whole
bias signal could be lost in the noise.
The issue is this, if you have a gross physical model of microsite bias, then you have a better chance of detecting that bias signal in a noisy enviroment. So before you try to detect the microsite bias it might be a good thing to have some sort of hypothesis about its phenomenlogy. How will this bias present itself?
So a physical understanding makes statistical testing more robust.
Of course the CRN site had the answer on the IR sensor:
‘This temperature is measured to determine the effective “skin temperature” of the “field of view” ground surface. Ground Surface (Skin) Temperature, along with wind speed and solar radiation provide information to allow for correction of observed air temperature data due to solar heating.’
So all the ground cover differences (which are considerable) go away.
Atmoz, IIRC there were many claims of great bias at the UofA station due to all the surrounding reflective surfaces. You point out that none is apparent in the comparison to the museum CRN station, so is this proof that the various data adjustments really have succeeded in eliminating the problem (step change issues aside)? I’m assuming that a bias must be apparent in the raw data.
[Reply: Actually, as I said in a previous post, there is a negative bias in the anomalies at the UofA station compared to surrounding stations.]
well ATmoz, hit anthonys site today, for a update on what
noa is doing to finish the CRN and update and modernize 1000
stations in the ushcn!
Steven, I agree with what you say. Anthony’s Stevenson Screen test is one example of what should be done to determine what the real affects of siting are. I was just curious to see if the differences I see in the graph wash out or not. Curiosity, nothing more. The CRN data time frame is just not long enough to cover site most moves or sensor changes that would cause discontinuities.
When the NCDC gets the CRN ftp site on line it will be easier to analyze how the sites might differ.
“now one thing I have been trying to get across to people is that Microsite bias is real, But it will only impact your trend estimate if you take a trend over the period in which the bias occurs.”
Why do you (still) have the try(!) to get this accross. Isn’t this fucking obvious? Do they have rocks in their heads? HAven’t we been throwing this whole issue around for several months? Wouldn’t they have learned this by now.
P.s. Tell Dave Smitth (who is rather decent) to use at least two sensors at his points A, B, and C. I was amazed that rocks in the head Watts gathered months of data on different painted samples and never bothered to consider innate sensor variability as a confounding factor. FYI2: the Dutch studt that Hans cites shows a quite good attempt to look at sensor to sensor differences (colocated, different instrument types). There is a photo there…and of course these guys have multiple sensors of the same type to check on innate within sensor class variation.
ignore spelling errors