U of Arizona Scientist Accuses NASA of Another Error

Pseudonymous U of Arizona climate scientist, Atmoz, has stated that the sharp increase of surface temperatures in 2000 in Lampasas TX recently reported at Climate Audit is not due to problems resulting from site relocation, but from an unidentified faulty algorithm (“UFA”) used by NASA.

Atmoz stated

Adjustments made by [NASA] GISS to this station seem to add an artifact that shows spurious warming since the year 2000.

He added:

The GISS temperature for this station has clearly “hockey-sticked” since the year 2000. I do not know the cause of this. But it is clearly a problem / feature of their algorithm and not an artifact in the data due to a station move.

“Amateur” reports on the Lampasas station had previously attributed the sudden increase in temperatures in 2000 to a station relocation from a park-like location to its present location surrounded by asphalt and a west-facing brick building. Atmoz rejects this explanation, saying that the amateurs failed to consider the possibility of a UFA.

Atmoz also reported the discovery of a USHCN station in Montana that is not located in a parking lot and that appears to comply with all WMO standards. Atmoz speculated that this news of contact with a USHCN station meeting all known WMO standards would not be reported at Climate Audit.

A Climate Audit spokesman said that they also welcomed the exciting news of contact and that they would immediately announce contact to their readers. He added that he was skeptical of Atmoz’ hypothesis that the Lampasas events were connected to another UFA at NASA, saying that other explanations still needed to be assessed.

A NOAA spokesman said that the agency had concerns about the news of contact and had already prepared contingency plans, including the potential relocation of the station to a nearby parking lot in Miles City, Montana shown in the picture below.

Atmoz is located in Tucson, Arizona, which pioneered the establishment of weather stations in parking lots long before Lampasas TX. Atmoz’ office overlooks the U of Arizona parking lot that hosts the weather station that has the highest temperature trend increase in the entire US Historical Climatology Network. Tucson is located in southern Arizona, approximately 350 miles to the west of Roswell, New Mexico.

48 Comments

  1. Gerald Machnee
    Posted Feb 17, 2008 at 12:40 AM | Permalink

    Has that parking lot been tested and verified as being up to standards, or was it chosen at random? We may get a complaint form the PLU. (parking lot union)

  2. Posted Feb 17, 2008 at 12:51 AM | Permalink

    I’ve been assured by two members of the United States Air Force that the apparently well-sited surface station was in fact an optical illusion caused when swamp gas from a weather balloon was trapped in a thermal pocket and then refracted the light from Venus, causing an apparent latitudinal shift in the station location. The station is actually located in a parking lot in downtown Miles City.

    Seriously, I rejected the station move hypothesis because a noticable temperature change was not seen in the raw data. Since a sudden jump in temperature was not seen in the raw or Filnet data, I assumed (perhaps wrongly) that it must be a problem with the NASS GISS algorithm.

  3. Max
    Posted Feb 17, 2008 at 12:53 AM | Permalink

    I would suspect it would need at least 2 air conditioners and one BBQ to fully comply. Standardization is a must in climatology.

  4. K
    Posted Feb 17, 2008 at 1:49 AM | Permalink

    I would differ from Atmoz about terms. UFA means Unidentified Fudging Algorithm or Undetected Fudging Algorithm. (the finest fudges are those undetected.)

    Faulty algorithms are mistakes. Fudging algorithms are not faulty. In fact they do exactly what is intended. (Any faulty fudging algorithm is quickly discarded because it does not produce the intended result.)

    However Atmos may be entirely correct in saying there is a Faulty Algorithm involved. Until we see the algorithm who can classify it as Faulty or Fudging?

  5. Stephen Richards
    Posted Feb 17, 2008 at 3:19 AM | Permalink

    UFA would surely affect all of the data and not just a single point. I could imagine that the repositioning might result in the graphs that both Steve and Anthony did but a UFA will require greater imagination.

    Eh! and by the way, I’m not keene on being called an amateur. I was paid to be a scientist so I think that disqualifies me. :)

  6. John A
    Posted Feb 17, 2008 at 6:25 AM | Permalink

    I have a modest proposal:

    Get rid of all of the rural stations. Simply setup lots and lots of stations in urban areas, on tarmac, near roads, airconditioning units and sun-facing brick walls.

    That way the USHCN will be calibrated to the same standard of incompetence that we have come to expect.

  7. John Lang
    Posted Feb 17, 2008 at 7:22 AM | Permalink

    Interesting that Atmoz managed to cut all his charts off right at the point where Lampasas’s temp records skyrocket off the charts due to the station move – leaving his readers with the impression that nothing is out of ordinary.

  8. Patrick M.
    Posted Feb 17, 2008 at 8:06 AM | Permalink

    I don’t know what the big deal is, the temperature station was teleconnected to the parking lot, (and other parking lots), long before it was ever moved from its previous park-like location.

  9. Mike B
    Posted Feb 17, 2008 at 8:25 AM | Permalink

    Interesting that Atmoz managed to cut all his charts off right at the point where Lampasas’s temp records skyrocket off the charts due to the station move – leaving his readers with the impression that nothing is out of ordinary.

    The temperature pattern at the two stations are nothing alike:

    Miles City
    Lampasas

    What the heck is Atmoz talking about? snip

  10. Jordan
    Posted Feb 17, 2008 at 9:19 AM | Permalink

    Interesting that Atmoz managed to cut all his charts off right at the point where Lampasas’s temp records skyrocket off the charts due to the station move – leaving his readers with the impression that nothing is out of ordinary.

    In technical terms, “Briffacation”.

    Steve: In fairness to Atmoz, he used a USHCN version that only goes to Dec 2005, although there is another version that goes into 2007. It’s not clear why there are different versions and it’s easy for someone unfamiliar with the data to get wrongfooted, so this is not Briffication sensu strictu.

  11. Anthony Watts
    Posted Feb 17, 2008 at 9:27 AM | Permalink

    Wow, this is deja vu all over again: Detroit Lakes.

    Wait until you see the “green” weather station I found. Still waiting on additional pictures of that one.

  12. Ron Cram
    Posted Feb 17, 2008 at 10:01 AM | Permalink

    re:2

    Atmoz wrote:

    Seriously, I rejected the station move hypothesis because a noticable temperature change was not seen in the raw data. Since a sudden jump in temperature was not seen in the raw or Filnet data, I assumed (perhaps wrongly) that it must be a problem with the NASS GISS algorithm.

    I do not understand how Atmoz can make this statement when Anthony showed the raw data in his original report and the raw data clearly spiked up.

    Atmoz, if you really want to defend this position – please show your temp chart and provide a link to your data set. Steve McIntyre says your data set ended in 2005. Could that be the problem? If it was a problem with the algorithm, wouldn’t the bias have happened all in one year?

  13. M. Jeff
    Posted Feb 17, 2008 at 10:11 AM | Permalink

    The asphalt doesn’t completely hide the crop circles.

  14. Mac Lorry
    Posted Feb 17, 2008 at 10:28 AM | Permalink

    How about doing some experimental science to resolve the controversy about whether or not the temperature is higher because of the new parking lot location for the Lampasas TX station compared to it’s prior location. This seems important enough given the number of moved stations to warrant an experiment such as putting in a second station in the original location of the Lampasas TX station. It would be a simple matter to then compare the temperatures and see what the difference is. If the same experiment was done on a number of moved stations scientists could better evaluate the temperature data for all parking lot stations.

    Without such a study I believe it’s proper to assume the temperature data from packing lot stations is biased beyond what any existing “correction” algorithm can account for.

  15. Demesure
    Posted Feb 17, 2008 at 10:34 AM | Permalink

    After all, UFA is just some letters away from UFO.

  16. Steve McIntyre
    Posted Feb 17, 2008 at 10:37 AM | Permalink

    #2. Atmoz, good for you in having a little fun with your response. There’s all too much unnecessary sourness in this debate and a little lightness now and then is healthy.

    I think that there is a noticeable difference in the USHCN data for Lampasas relative to nearby Blanco (see plot in comments on other thread from USHCN data – where I used the NOAA version which goes to 2007, not the CDIAC version which you used that ends in Dec 2005. Also I showed mean temperature while you showed max – I’m not sure why you showed max only. So I think that there is evidence of a change synchronous with the station move.

    Having said that, far be it from me to reject the possibility of another NASA UFA. I’ve been working though the GISS versions this morning, comparing the present online versions to ones that I scraped last August, and there are some curious differences that I’m going to write up. It gets a little hard to keep up with all the NASA versions sometimes.

  17. Doug
    Posted Feb 17, 2008 at 10:42 AM | Permalink

    We have to define “park like setting”.
    I have been looking at houses, and the experts in the field, known as Realtors, seem to need a Watts style audit as to what they deem to be “a park like setting” I think the current location actually qualifies.

  18. Mike B
    Posted Feb 17, 2008 at 10:44 AM | Permalink

    Steve: In fairness to Atmoz, he used a USHCN version that only goes to Dec 2005, although there is another version that goes into 2007. It’s not clear why there are different versions and it’s easy for someone unfamiliar with the data to get wrongfooted, so this is not Briffication sensu strictu.

    Since his post clearly refers to GISS, why didn’t he use the GISS website for the data? “Out of fairness” to Atmoz, I guess one could assume he doesn’t know the difference between NOAA and NASA.

    Or alternatively, “out of fairness” to Anthony Watts…

  19. kim
    Posted Feb 17, 2008 at 10:56 AM | Permalink

    Doug, #17, there are clearly parked cars, setting.
    ==============================

  20. Posted Feb 17, 2008 at 11:53 AM | Permalink

    The data I used does end in 2005. I obtained it here. I don’t know of another location where it is possible to download the entire USHCN dataset through 2007, but would gladly update my plots if pointed to such a location.

    I used the max temperature because the mean includes changes to the max and the min (obviously) and is thus more difficult to interprete. I did look at the mean temperature through 2007, but did not see anything that was noticably different than the max plot I showed.

  21. steven mosher
    Posted Feb 17, 2008 at 12:12 PM | Permalink

    RE 16 Atmoz has been quite cordial. He also has done some nice posts on the oversampling
    of the US. Everytime I try to make this srgument at other places ( arguing for using quality stations)
    people just react in kneejerk fashion.

    There was also a paper on this published in 2004.. I linked it around here somewhere.
    ( shuffles through his papers)

    I’m also digging back into GISS code.. SteveMC at some point Hansen splits Nighlights
    into three classes.. Unlit, dim, bright. Do you recall where he does that or what
    breakpoints he uses

  22. Steve McIntyre
    Posted Feb 17, 2008 at 1:39 PM | Permalink

    #21. in Anthony Watts’ very first post on stations comparing Orland and Marysville, Anthony observed that Orland was a good station and Marysville a bad station. The existence of “good stations” was noted right from the outset.

    NOAA deserves no credit for having a “good station” at Miles City – that’s what they’re supposed to have. The LAmpasas case is especially objectionable because of the deterioration of station quality on Tom Karl’s watch.

  23. David_Jay
    Posted Feb 17, 2008 at 2:15 PM | Permalink

    I have been looking at houses, and the experts in the field, known as Realtors, seem to need a Watts style audit as to what they deem to be “a park like setting”

    According to realtors, all homes would qualify (at something under 1 Sigma). I think this calls for an engineering-quality study!

  24. RCB
    Posted Feb 17, 2008 at 4:48 PM | Permalink

    Steve, I like your reference to Roswell, NM – the implications are limitless!

    Still very much looking forward to your detailed comments from your experiences in
    “Off to Georgia Tech”.

  25. kim
    Posted Feb 17, 2008 at 4:54 PM | Permalink

    I’m curious, too, but curiosity killed the cat. Maybe what happened in Georgia should stay in Georgia. It might be difficult to satisfy all the critics.
    =============

  26. LadyGray
    Posted Feb 17, 2008 at 5:44 PM | Permalink

    The entire quote is:

    Curiosity killed the cat. Satisfaction brought him back.

  27. kim
    Posted Feb 17, 2008 at 5:46 PM | Permalink

    Excellent, then, it might be difficult to satisfy all the cats.
    =================================

  28. John Andrews
    Posted Feb 17, 2008 at 6:44 PM | Permalink

    Toward the end of March, I will become very cautious about what I believe and do not believe in both the main report and the comments section of this blog. There is too much talent on exhibit here to not do so.

  29. Posted Feb 17, 2008 at 6:56 PM | Permalink

    Doug (#17) writes,

    We have to define “park like setting”.

    It’s where you can like, park.

  30. yorick
    Posted Feb 17, 2008 at 8:22 PM | Permalink

    That’s right, it’s British English, Car Park-like setting.

  31. Ron Cram
    Posted Feb 17, 2008 at 9:50 PM | Permalink

    re: 20
    Atmoz,

    You write:

    I used the max temperature because the mean includes changes to the max and the min (obviously) and is thus more difficult to interprete. I did look at the mean temperature through 2007, but did not see anything that was noticably different than the max plot I showed.

    I am still baffled. Anthony Watts provided the raw data through 2007 in his first post and the sudden increase beginning in 2000 was striking. He also compared this data to the GISS post-homogeneity adjustment. The adjustment made the past cooler but did nothing to cool the station after its move to the parking lot. Did you find something wrong with the data Anthony provided?

    Also, I’m confused on another point. You say you looked at the mean temp data through 2007 but did not see anything different from the truncated Tmax dataset you referenced. How can that be? Didn’t Anthony use the mean temp data through 2007?

  32. MarkW
    Posted Feb 18, 2008 at 7:02 AM | Permalink

    Cats are notoriously difficult to satsify.

    Regardless, it’s Georgia Tech, Georgia is that other school.

  33. Posted Feb 18, 2008 at 10:06 AM | Permalink

    Ron Cram: Anthony showed only GISS data – pre-adjusted and post-adjusted. I showed the ‘raw’ and ‘filnet’ data from USHCN. They do some processing of their own on the data, including adjusting for outliers (raw), time of observation bias (TOBS), adjustments for the bias of the MMTS (MMTS), station moves (SHAP), and for missing values (FILNET). They also offer a seperate product that attempts to account for the urban heat island.

    Steven Mosher has been kind enough to read through the Hansen (2001) paper and figure out that they use the data just before the FILNET correction. This means that before NASA GISS even get the data there is supposedly already a correction in place for the station move. Therefore, if you wish to see the effects of the station move in the temperature record, you should not be looking at the GISS data. You need to look at the data pre-SHAP adjustment, which I provided.

    Steve: Atmoz, you have to be a bit careful here as GISS handling of USHCN data has been in flux this year. Prior to August, they seem to have used USHCN SHAP version up to 2000 and then GHCN for 2000 on – yielding the “Y2K” problem as there were problems with the splice that were well-publicized. In mid-August, they re-issued their data making an ad hoc patch for each series to eliminate Y2K discontinuities on a station by station basis. Then in mid-September, they changed again – and this was a very annoying change – this time substituting the FILNET version (without interpolated values) for the SHAP version, creating a whole lot of peculiar differences with their prior values. I just re-visited their site and, for the first time, they’ve put a little effort into identifying the exact data sets that they are presently using. But you can’t take the descriptions in H2001 as definitive.

  34. Mike B
    Posted Feb 18, 2008 at 11:11 AM | Permalink

    Steven Mosher has been kind enough to read through the Hansen (2001) paper and figure out that they use the data just before the FILNET correction. This means that before NASA GISS even get the data there is supposedly already a correction in place for the station move. Therefore, if you wish to see the effects of the station move in the temperature record, you should not be looking at the GISS data. You need to look at the data pre-SHAP adjustment, which I provided.

    An attempt by NOAA to adjust for the station relocation does not guarantee an adequate adjustment.

    At the end of the day, the facts remain:
    Lampasas is a GISS “sweet spot” for warming, unlike its neighbors, rural or urban.
    The current location of the station that generates the data is about as bad as you can get.
    Lampasas became a sweet at about spot about the time it was relocated.
    Miles City, Montana has nothing to do with any of the above.

  35. Sam Urbinto
    Posted Feb 18, 2008 at 11:56 AM | Permalink

    Badly sited, well sited, it’s about finding what’s good and what’s not. While it would be freferable to have good stations and not have to try and adjust for issues, there is what there is. The question is though, if the status of the sation is not know, how can it be sure it’s being adjusted accurately. This is of course aside from the fact that these stations are placed according to weather reasons by convenience, rather that specifically to track climate. And as I’ve said, proxying the area’s temperature by sampling air near the ground in sparse locations tens or hundreds of kilometers from each other and gathering the mean of min/max.

    Hardly anything to get angry about.

  36. steven mosher
    Posted Feb 18, 2008 at 12:48 PM | Permalink

    RE 33. SteveMc. I referred ATMOZ to H2001 and the line of Fortran in GISSTemp which reads in NoFill.

    A while back I think some of us suggested a flow diagram of sorts to show where all this
    stuff comes from.. picture 1K words…

    All that said, Version2 of USHCN is going to make scrambled eggs of all of this.
    From the way it is described a single program, a change point analysis program, will
    account for all documented and UNDOCUMENTED stations changes.

    Without a description of this program we are back at square zero.

    With SHAP ( which must adjust for latitude and elevation at least ) you could double
    check or at least make some sense of it.

    Maybe somebody can free up the USCHN papers behind the green Wall.

    Totally unrelated, there is a CRN paper on station density requirements ( Vose I believe)
    that purports to demonstrate that for the US only 135 stations are needed for climate
    Trend data. I linked it here a while back.

  37. steven mosher
    Posted Feb 18, 2008 at 1:20 PM | Permalink

    RE 34. MikeB

    This entire adjustment effort is a clown parade in a culdasac: A clusterfun.
    ( hehe you know what I really meant)

    Start with TOBS. time of observation
    Bias. Now, to be sure, if you observe min/max at different times, you get different
    results. The paper establishing the TOBS correction was written 20 years ago. The std
    error in the model was .2C. TOBS adjustments warmed the US. they have not been audited.

    MMTS adjustments, when NOAA switched to MMTS quayle argued that the switch COOLED the record.
    so, we get MMTS adjustment. A 20 year study of MMTS sited side by side with the Historical
    systems shows that Quayle(1991) may have over estimated the cooling. But that adjustment goes in.

    SHAP: I dont have the SHAP code although Peterson says its available. Given his other writings,
    and given the data in the station history file One can assume that he adjusts for LAPSE RATE and
    Latitude. That is, if station A moves from 0 feet above sea level to 300 feet, then a Lapse
    rate change will apply… same for latitude changes.. ( speculation ok?)

  38. Mike B
    Posted Feb 18, 2008 at 1:48 PM | Permalink

    This entire adjustment effort is a clown parade in a culdasac: A clusterfun.
    ( hehe you know what I really meant)

    Absolutely agree.

    IIRC, about half of the warming since the mid-70’s comes directly from TOBS. There is no question that time of observation affects the daily min/max average. Can’t find a linky though.

  39. Mike B
    Posted Feb 18, 2008 at 1:52 PM | Permalink

    #38 didn’t come through quite right…meant to say that there is potentially a bias from time of observation. However I’m unconvinced that there is compelling evidence that there has been a systematic change in the network-wide observation pattern that requires such a dramatic adjustment.

  40. Sam Urbinto
    Posted Feb 18, 2008 at 4:41 PM | Permalink

    What would make anyone think that getting the mean daily temperature is particularly helpful in telling us what “the temperature” is? So min or max or both are going up (perhaps; it’s after massaging the data) over time. .6 C trend over 125 years? OMG OMG OMG

    sheesh

  41. Andrew
    Posted Feb 18, 2008 at 6:54 PM | Permalink

    Is Atmoz going to take back the title of his post (ie, that you would “never see” the station “profiled” here)?

  42. Posted Feb 19, 2008 at 10:08 AM | Permalink

    Tucson is located in southern Arizona, approximately 350 miles to the west of Roswell, New Mexico.

    Cracked me up.

  43. Dave
    Posted Feb 19, 2008 at 12:50 PM | Permalink

    From http://data.giss.nasa.gov/work/gistemp/STATIONS//tmp.425722570030.1.1/station.txt

    YEAR JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
    1996 6.4 10.6 10.5 17.2 25.5 26.7 29.5 27.6 23.2 18.1 13.0 8.7
    1997 6.8 8.5 13.5 14.7 20.3 24.8 27.9 27.5 25.7 19.1 10.6 6.3
    1998 9.6 9.0 11.9 16.0 24.4 28.1 29.6 27.3 25.7 19.5 14.1 8.1
    1999 9.6 11.9 13.0 18.9 21.9 26.1 27.5 29.2 24.7 18.3 14.1 8.2
    2000 9.0 12.2 15.3 19.7 24.2 999.9 999.9 999.9 999.9 999.9 11.0 5.6
    2001 6.8 11.2 10.7 19.9 23.8 27.2 29.8 999.9 24.6 19.1 16.1 9.8
    2002 9.6 8.4 12.5 22.7 24.0 27.1 27.6 29.3 25.7 18.5 11.8 9.3
    2003 7.3 7.8 13.2 19.5 25.4 26.1 999.9 29.8 999.9 21.0 999.9 9.7
    2004 10.2 8.7 16.7 18.8 23.1 26.3 999.9 27.6 24.8 23.0 15.2 10.0
    2005 999.9 10.9 999.9 999.9 22.5 28.3 29.7 28.5 28.4 999.9 16.7 999.9
    2006 999.9 9.8 15.2 999.9 999.9 999.9 999.9 999.9 999.9 999.9 999.9 999.9

    I guess I’d look out of whack with numbers like this….what’s with the 999.9 all over the place?
    A number of other “Hot” years had the same thing in them included 1892-1894.

    Steve:
    Those are code numbers for NA. They are excluded from calcs and do not cause a problem.

  44. Dave
    Posted Feb 19, 2008 at 3:04 PM | Permalink

    Steve,

    I realized after that 999.9 meant that it was excluded.

    My point is that it looks like missing data is corrupting ANN (Last column) for Lampasas.

    It looks like ANN = Average (D-J-F,M-A-M,J-J-A,S-O-N). A simple calculation of complete years seems to bear this out.

    For 2004 (9.5 + 19.5 + 27.3 + 21 )/4 = 19.325
    For 2006 (10 + 28.8 + 22.6)/3 = 20.467 slightly higher than 20.45 posted

    Obviously they have to do something to account for this missing data (for 2006 M-A-M)? The question is how do they do it?

    For Lampasas, there is 4 missing periods from 2000 – 2005 and 12 missing months of data when the jump in temperature occurs.

  45. Paul Foote
    Posted Feb 21, 2008 at 12:32 PM | Permalink

    Atmoz,

    I read the post and responses on your website that started this post, and I think your assertion that “A Surface Station You’ll Never See Profiled at Climate Audit” can be shown as untrue. Considering the polite response you have had from Steve, Andrew Watts and Stephen Mosher, I request that you include ClimateAudit.org on your blogroll.

    I also think you should ammend the title of your post on your site. I’ve never found Steve to intentionally shun any “climate science” controversy as proven by this post and many others.

  46. Posted Feb 21, 2008 at 4:34 PM | Permalink

    The title was written in jest in an attempt to get Steve to link to my post. Given the nature of Steve’s reply here, I did not think he was offended by the title. I will not change the title (because of the way WordPress handles its pretty URLs), but I would add a qualifier to the post if asked to by Steve.

  47. Steve McIntyre
    Posted Feb 21, 2008 at 4:53 PM | Permalink

    #46. It’s pretty small beer in terms of offences.

    Orland was recognized as being a “good” site in the very first post on this matter.

    Obviously “bad” sites tend to attract more attention than “good” sites, but the underlying question has been to establish some objective meta data and then see whether there’s any differences. In the Peterson data set, more precise demarcation of sites led to distinct patterns between urban and rural sites, which were jumbled in his classification.

    My own surmise, as I mentioned last fall, is that the NASA US version is more similar to results from “good” stations than the NOAA version – which is worth knowing. Also that there’s enough difference between “good” and “bad” sites in the US to warrant evaluation of ROW sites and, in particular, the effectiveness of the R-classification in replicating objective circumstances at “gGood” US sites.

  48. steven mosher
    Posted Feb 22, 2008 at 8:14 AM | Permalink

    I want to thank Atmoz for highlighting the miles city site. It led us to look afresh at the
    Rural/smalltown/Urban classification and the nighlights classification and improve the metadata
    that Anthony has been putting online. The 250 or so sites that GISS actually uses for the US
    are now identified. So, its time to redo some of the OPentemp work that JOHNV started.

Follow

Get every new post delivered to your Inbox.

Join 3,191 other followers

%d bloggers like this: