Feb 16 2008
GISS Adjustments to Miles City - USHCN#245690
Expanding upon a previous post about the surface station at Miles City, Montana, this entry will again look at that station, but will show what NASA GISS does to this data in its adjustments. A quick summary of the last post: I wanted to find a site that USHCN had classified as not being influenced by humans. They provide a product that describes the area around each surface station, and the Miles City location was classified as open farmland, grassland, or tundra out to at least 10 kilometers. One of the original reasons for looking at this station was because I wanted to look at the data from a ‘good’ station that had been profiled by surfacestations.org. A sample photo is provided below.
|
Surface station photo courtesy of Anthony Watts, www.surfacestations.org and Craig Limesand.
Steven Mosher commented on the previous post about the Miles City station that it might be interesting to look at the data provided by GISS to see what they do during their adjustments.
Miles city is Rural as you note. So, do you think it would OK to adjust its temperature UPWARD based on comparions with nearby sites? It’s class 1. Its Rural. What would you bet? It’s rural. It meets standards, you would think that other sites adjust to it. You wouldnt adjust this site one way or the other. You’d adjust the bad sites.. right?
Miles City is rural, I think it’s hard to disagree with that. There should be no reason to expect that this station would need adjustments due to the urban heat island (UHI). However, I would bet that most of Montana is rural. And the purpose of the adjustments is not just to correct for UHI, but to account for other inhomogeneities as well. While I have downloaded the GISTemp code, I find it harder to read than my code (mostly because Fortran sucks).
Therefore, I would not expect that the Miles City data would not be adjusted, but I would expect that it would not be adjusted much. Obvious reasons for adjustments unrelated to UHI include station location moves, urban encroachment (not an issue here since the last station move), and instrument changes. [Up to this point was written before looking at the data.]
|
The above plot shows the data that GISS obtained from the USHCN (in black) and the data after their homogenization efforts. As can be seen, there is a small difference between the two, so GISS has adjusted the data for some reason. Also note that the adjusted values are always greater than or equal to the raw data. The following two plots show the raw data (top) and adjusted (bottom) and their associated trends.
|
|
The adjustments that NASA GISS apply to the USHCN data increase the linear temperature trend since the beginning of time series. However, this change in the trend is well within the error of the measurements. More information can be gained by looking at the difference between the two time series. The plot below shows the GISS adjusted values subtracted from the raw values provided by the USHCN.
|
The bulk of the adjustments occured during the 1960s-1980s when GISS adjusted this stations values upward by 0.3C. Since that time, the adjustments have decreased, and there has been no adjustments to this stations data since 2001. What is interesting about this graph is that the GISS adjustments since 1980 have decreased the temperature trend. Therefore, the difference in the warming trends is mostly due to the corrections from before the 1950s. It is these adjustments that cause the greater warming trend in the adjusted data as compared to the raw data.
Conclusions
For the last several years, GISS has not adjusted the temperature data from the USHCN. In the past, they had increased its value by up to 0.3C. I am unsure why they have done this. I am unsure of the exact reason for their correction. However, the homogenization techniques could also try to adjust for spatial changes in temperatures.
For instance, suppose there are three stations located all located at the same latitude. The first station is 400 kilometers from the second, and the third is 400 kilometers from the second and 800 kilometers from the first. If the first station has an average monthly temperature of 10.3C and the third station 10.7C it would be reasonable to assume that the temperature at station two would be 10.5C (all else being equal). If the monthly temperature for station two was 10.4, it might reasonably be adjusted upward by 0.1C.
(No Ratings Yet)
Loading ...Related Posts:
25 Responses to “GISS Adjustments to Miles City - USHCN#245690”
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.
Your analysis has a flaw. ASOS sites are new sites. Obviously this was not an ASOS site in 1900. The station history shows at least three locations: Miles City (til 1938), Miles City FAA Airport (til 1990) and the Miles City FCWOS. There have also been a number of changes in between (I don’t have the history file format available so I can’t say what they are). So the only part of the temperature history that you can say with any certainty is a class 1 is the time from 1991.
The station history (4.6Mb) file shows that the last station move occured January 1, 1938. Entries after that are for other changes, such as instrumentation. To decode the station history file, see the documentation.
http://www4.ncdc.noaa.gov/cgi-win/wwcgi.dll?wwDI~StnSrch~StnID~20012408
Note move in 1996 and changes in elevation which can effect the data also.
That “move” is almost certainly due to this. Even if it weren’t, the station would have only moved 1 minute of longitude, which at this particular latitude is about 1/2 mile. I have previously shown that surface temperatures are highly correlated at distances of at least 250 kilometers (~150 miles). A change of 1/2 mile would be insignificant.
The change is height given is 0.3 meters. Assuming a lapse rate of 9.8 C/km, that would introduce a temperature change of 0.003C. I think we’re safe on this one.
It is interesting that the NOAA site has a different history than CDIAC.
Atmoz,
While it is possible to find examples showing that “surface temperatures are highly correlated at distances of at least 250 kilometers”, it is hardly valid that you’ve shown this to be true in general.
I’ve had a running conversation with a friend who works at our local airport and helps manage abatement of weather induced effects on landings/takeoffs (ice, etc on runways). They have temperature sensors at many points around the airport and nearby areas. They frequently find temperature differences of several degrees over the length of a runway, and even more (up to 5-6 degrees) when comparing with sensors up to three miles away.
Gardeners are typically aware of these kinds of microclimate differences and their impact on plant growth. My wife is a Colorado Master Gardener. Along the front range of Colorado (with similar altitudes) you can find a consistent variation of several weeks in the typical bloom date of various plants. This of course is due to a variety of effects, but temperature is one of them. The effects are recorded in detail through Sunset Magazine’s Climate Zone maps. While the small scale maps (such as here) suggest a broad uniformity, that is simply due to the lack of available detail.
Find a copy of the Sunset Western Garden book. Examine the gardening zone maps for the San Francisco Bay Area, which is local to the magazine and where they have the most experience. You’ll find detailed maps showing some tremendous climate variation at the neighborhood level. This was an invaluable and practical resource to us when we lived there.
To reiterate my bottom line: climate varies tremendously, even over short distances, in practical ways that are important to many people. It’s hardly valid to claim uniformity across the board.
Computer code languages don’t write bad code, people write bad code.
And many of us are happy to see others independently come to the same conclusions regarding GISS/NASA coding practices ( of which there doesn’t seem to be any ).
“However, this change in the trend is well within the error of the measurements.”
Does a horizontal line fall within the error of the measurements? BTW, where are the error ranges on those fits and the measurements?
Atmoz thanks.
station move adjustments are done prior to giss ingesting the data. ushcn SHAP takes care of that. i’ll go back to the code and have a look, also hansens papers
It’s worth noting that since the move to the rural site in 1996, Miles City has been seeing a cooling trend.
Whether stations are highly correlated within 250 km (and I agree it’s true in general) is an entirely separate issue from whether station moves introduce an inhomogeneity. Consider a simple idealized thought experiment: two nearby locations, one of which (because of microclimate) systematically differs from the other by roughly 0.5 C. The records of annual temperatures from the two sites would be highly correlated, but if a station was located in one site for the first half of the century and the other site for the second half, a significant trend error would be introduced that would need to be corrected for somehow.
One’s intuition might say that the erroneous trend would have a magnitude of 0.5 C per century, but if you do the math, the trend error is actually much worse: 1.0 C per century.
Before assuming that small station moves over relatively flat terrain don’t make much difference in the temperature measurements, consider the experience of Reno (scroll down halfway).
The Reno link is http://www.weather.gov/om/csd/pds/PCU6/IC6_2/tutorial1/Impacts.htm.
BTW, I took a look at the current Sunset Western Garden Book maps. Fascinating updates!
For example, in Anchorage AK, they show how neighborhood A may be as much as 25F warmer than neighborhood B in winter, and B as much as 10F warmer than A in summer. All due to microclimates, and not UHI.
Pretty crazy stuff!
hi I just checked hansen 2001 and all station move adustments
are made by uhsch, then giss ingests, then giss does a Urban adjustment. adjusting urban and small town ( pop >10000) to
surrounding rural sites. Miles city has a pop < 10000 and has had
a population this small according to us census figures going back to
the start of the temp record. I sure there is an explnation, im probably missing something
I figured it out.
In hansen 2001 Hansen says he uses nightlights to determine
if a station is Rural in the US and population everywhere else.
Miles city population is less than 10K which makes it rural,
BUT, nightlights ( satillite imagery taken in 1995) indicates a brightness factor for Miles of 26! effectively making it urban.
For a contrast the city of Susanville CA, has nightlights of 0!
population 13,000
Google map it
For constrast Quincy california has a nightslights of 12
Population 1879
Nightlights might have some issues in identifying rural.
I realize the actual change in altitude is tiny with the lapse rate. I wasn’t clear that what I was referring to was a possible change in height of the sensor above the ground which has a larger effect, not to mention possible changes in the type of sensor used.
The correlation with distance is only valid when the two sites maintain consistent environments. If both are in fields and one is moved to a concrete pad that will change the relationship.
BarryW and others.
Nasa GISS ingest data from GHCN AFTER adjustments are made.
THEN GISS make Urban adjustments.
USHCN makes these adjustments
1. TOBS.
2. MMTS
3. SHAP.
4. FILNET
GISS Read in the data BEFORE the FILNET adjustment.
open(10,form=’formatted’,file=’USHCN.v2.mean_noFIL’)
The USHCN adjustments are decribed in detail at their site.
in short.
1. TOBS: modifies data based on a model to adjust for differences
in time of observation ( when the thermometer was read)
This model was written in the 1980s by Tom Karl. The std error of the model was .2C I believe. TOBS is a sizeable adjustment. It warms the record.
2. MMTS. You have all seen the square little boxes called Stevenson Shelters. These have been switching over time
to a device called the MMTS ( looks like a beehive) USHCN, based
on a study by Quayle in 1991, decided that these devices were registering too cool. So, they adjust the non MMTS stations.
Quayle study may have some issues, as noted in this 20 year study
with side by side instruments
http://ams.confex.com/ams/pdfpapers/91613.pdf
3. SHAP. SHAP adjusts stations for changes in instrumentation and location. For example changes in latitude or elevation.
4. FILNET. GISS do their own Infilling of missing data. Mistakes in this have been Identified at CA by several folks, John Goetz in particualr.
The current approach at USHCN differs slightly
http://www.ncdc.noaa.gov/oa/climate/research/ushcn/
So, GISS ingests USHCN data AFTER adjustments have been made
for Time of observation changes, Instrument changes, and Station move changes.
Then, they do their own INFILL of data and then they do a Homogenization adjustment. In this stage they Adjust Urban sites
to “match” rural sites.
In the US ( and Canada I think) Rural is defined by a ( i think) 1995 satillite photo of lights at night. This snapshot in 1995 determines whether the site was rural in 1930 ( just kidding but that is what it does in this method)
For the Rest of the World Rural is determined by Population.
I believe the population data is taken from 1980 or so.
So, You dont visit a site to determine it is rural. You look at old population data or old satillite data.
For grins Google Mineral, California. This town of about 150 people
is not considered Rural by nightlights.
In short, Rural is determined by PROXIES. 1 population figure
and 1 satillite photo.
BarryW: The correlation with distance is going to be relatively valid for the distances we’re talking about with this station.
Steven Mosher: Thanks for figuring out what GISS is doing to this station. Where do you find out the value of “nightlights”? I’ve wanted to look at the difference between adjusting based on lights and population.
MrPete: In this post I wasn’t trying to generalize to every station. The distances moved by this station since 1937 are essentially nil, and as I said above are most likely due to someone going to the station with a GPS and getting more accurate coordinates.
@papertiger: There was no station move in 1996.
@Dan Hughes: Since I write crappy code I prefer to blame it on the language.
The errors in the trends are shown on the plots. A zero-trend level is not within the (one-sigma) error level. The USHCN data do not come with error estimates (that I know of), so when calculating the error in the trends I assumed that the error in the temperatures was small.
@Vic Sage: I plotted all the data that was available in the version of the USHCN available here.
Nightlights is here:
http://data.giss.nasa.gov/gistemp/station_data/station_list.txt
Its the last column marked brightness index. I Think Dr. Hansen started using this in 2001 ( possibly before) I’m rereading the papers.
In the GISS code the brightness index is reduced to three categories . Unlit dim bright, corresponding to Rural, periurban, and
Urban. I havent found the cut off points. Hansen 2001 has a table I believe.
I believe the satillite photo is circa 95 ( I’m doing at lot from memory sorry) and I beleive the pixel extent was something like 2.6km???? ( that sounds wrong to me, I’ll check)
Oddly, the few sites I have visited ( orland CA. Susanville, CA.
Quincy CA. and Mineral CA. ) were all a bit off. The urban
( susanville) coming in at unlit and quincy, mineral and orland coming in a peri urban or urban.
Odd, to say the least.
The station list above is the Current one posted at nasa. I suggest a check of the latest code to see if it has changed.
I will also point you at the fortran that does the adjustment.
Its rather challenging.
Atmoz..
The urban adjustment happens in Step 2 of the Giss Code.
( you have it right?)
The routine is PApars.f
Actually, If you have a smart student it would cool to get the code
compiled and working. After a couple weeks of working at this
several guys gave up. We are pretty sure they used AIX.
The red headed step child of unix.
Nevertheless, We worked through a bunch of minor IO problems and compiler switch issues
but we could never get the entire thing running and then we got ADD and moved on.
I think it would be an instructive excercise for students to read and understand OPC. other peoples code. even when that code is,
how shall I say it politely, not exactly clear. Some people thrive
on this. In the end, the students will learn that other people need to be able to:
1. read their work
2. See all the steps.
3. repeat all the steps
4. Improve the method
5. Maintain the method.
I’ll just end this by saying I like your blog and your good sense of humor and your willingness to look at things.
It’s not the distance it’s the movement of the station to a new microsite. Think of the effect of moving the temp gage from the north side of a building to the south side. A quarter of a mile can make a big difference in the microsite= from next to a building to a field for example.
Atmoz, the cutoffs ( categories) for Rural/periurban/Urban are
in hansen 2001. Miles city would be categorized as Periurban
( small city) and would be adjusticated accordingly.
@papertiger: There was no station move in 1996.
The station curator’s notes in the site survey refer to two moves. One in ‘96 and another in 2000.
I quote “Same location since 2000..”
Why would someone say that if there had been no station move?
I agree that the station may not have moved at all. However, your argument was that a 1/2 mile change would be insignificant in light of the correlation you found between temperatures 250km apart:
“I have previously shown that surface temperatures are highly correlated at distances of at least 250 kilometers (~150 miles). A change of 1/2 mile would be insignificant.”
What I’m providing is evidence that, if the move were even as small as 1/2 mile, the change could be quite significant.
Here’s another example: the morning temperature behind my home tends to be much colder than in front, because there’s a substantial slope behind my house — with a large amount of downslope cold air that “drains” down toward my house. The house blocks the cold air mass from directly moving to the front. 40m horizontal difference in location, zero altitude change, yet several degree difference visible in frost patterns, etc.
You really have to look at the lay of the land, locally, to have any confidence in microclimate issues.
I just wanted to clarify my thoughts on this issue. Note this is specific to the Miles City station.
I don’t think there was an actual station move since 1937. The change in the location in the meta data was probably because someone went to the station with a GPS. The station has been located at the airport since at least 1937. Even if the station did move, the distance would likely be better measured with feet than miles. Small airports, such as at Miles City, are usually flat and homogenous (except the runways and taxiways); shown nicely by the photographs at surfacestations.org.
Any station moves would likely be at the same elevation and very similar land cover. Therefore, microclimate issues at this station since 1937 are likely to be small.
I’ll make one more comment before I drop the subject. If you look at the details in the MMS system for the last three dates you’ll see one with zeroes in the seconds column and two with values. I would guess the last two are both GPS values and do show a small move. I agree with your opinion that the shift from zeroes to a seconds value it probably just better reporting of the location. The last two in my opinion probably show the installation of the ASOS that we see in the surfacestation photos but I’m just guessing. What effect that may have on the data I think is where we differ, since we don’t know what the CRN classification would have been of the previous site (e.g., parking lot as opposed to a field).
The Elephant in the Room is that GISS rate Miles city as
BRIGHT.
H2001 references Imohoff, where BRIGHT was defined by characterizing 3 urban cities: Miami, Sacramento, and Chicago.
These cities characterize the BRIGHT category and establish that categories correlation to Urbanization.
If anyone out there believe that Miles city is in the same class as
Chicago, I have some real estate on the Miracle mile to sell you.
Again, These potential weirdnesses do not mean there is no global warming. They just mean some more attention could be paid to certain details. Paying attention to those details is a joy for some people.
More later too little time too many papers.