by John Goetz
Update: Thanks to an email from John S. – a patron of climateaudit.org – we have learned that the Russian data in NOAA’s GHCN v2.mean dataset is corrupted. For most (if not all) stations in Russia, the September data has been replicated as October data, artificially raising the October temperature many degrees. The data from NOAA is used by GISS to calculate the global temperature. Thus the record-setting anomaly for October 2008 is invalid and we await the highly-publicised corrections from NOAA and GISS.
Update 2: The faulty results have been (mostly) backed out of the GISS website. The rest should be done following the federal holiday. GISS says they will update the analysis once they confirm with NOAA that the software problems have been corrected. I also removed the subtitles since the GISS data no longer reflects October as being the warmest ever.
GISS (Goddard Institute of Space Studies) Surface Temperature Analysis (GISSTemp) released their monthly global temperature anomaly data for October 2008. Following is the monthly global ∆T from January to October 2008:
Year J F M A M J J A S O
2007 85 61 59 64 55 53 53 56 50 54
2008 14 25 62 36 40 32 52 39 50 78
Here is a plot of the GISSTemp monthly anomaly since January 1979 (keeping in line with the time period displayed for UAH). I have added a simple 12-month moving average displayed in red.
The addition of October has changed some of the temperatures for earlier months:
GISS 2008 J F M A M J J A S O
As of 9/08 14 25 62 36 40 29 53 50 49 ..
As of 10/08 14 25 62 36 40 32 52 39 50 78
The 0.78 C anomaly in October is the largest ever for October, and one of the largest anomalies ever recorded. Although North America was cooler than normal, Asia apparently suffered from a massive heat wave.
Also, after several months of being downgraded to a 0.61 C anomaly, 2005 has been lifted back to 0.62 C.
Sponsored IT training links:
Enjoy the first hand success with 646-046 online training. This all in one 642-975 training package includes everything you need to pass 000-106 exam.
Steven Hill (05:54:06) Southeastern sky, my good man, southeast.
====================================
From NOAA website:
Contacts
For questions/assistance with GHCN-Monthly data, please contact Russell.Vose@noaa.gov
For assistance with the GHCN-Monthly web site, please contact Jon.Burroughs@noaa.gov
Not to be a pest, but someone last month had noticed that the Finnish Sept data were duplicates of data from August. I just rechecked, and these data are still in the Oct. dataset
Oulu
Sept. 4 WU 7.4 GISS
Aug.
Interestingly, there are no actual data for Oct. Apparently, someone saw something was odd, but they didn’t bother to change the previous data.
The problem with September data replicated as October data also applies to all stations in Finland.
Sorry about the last post. I just rechecked, and these data have been duplicated in Oct.:
Oulu
Oct. 4 WU 7.4 GISS
Sept. 6 WU 7.4 GISS
And the WeatherUnderground means are lower than the “real” GISS data.
Chris V
I hope this “mistake” will soon and publicly be corrected, I personally doubt it.
GISS takes a lot of flack not because of this error but because it is widely reported as authoritative and has a history of very questionable quality. They have more than 70 “corrections” of their past record in the last few years these changes overwelmingly make the past cooler and the present warmer. Documentation for these changes range from doubtful to non existent but they are much of the supposed warming signal. We have proof that the USHCN that is a part of GISS is strongly biased toward warming and that the bias has not been accounted for. No corrections or explanations are forthcoming on these problems, so the hope they will correct or even acknowledge this one is overly optimistic.
This is not an unheard of blog, they almost certainly know that their October numbers in Siberia are silly, have they made a correction or issued a warning? Wouldn’t you have done so if the work was yours?
[…] Chris commenting at WUWT noticed that some of the GISS temperature data records for Northern Asia at GISS show identical readings for September and October. For example: Turuhansk (65.8 N, 87.9E) which whows 8.1 C for both September and October. […]
JohnGoetz:
No. If GISS’s temperature is affected (as it appears to be), the error is shared. It’s true the GISS gets their data from NOAA. So, if NOAA makes an error, it will propagate.
That said, GISS puts out a data product. They are responsible for their product.
The process involves humans who should have written a script to check for certain types of outliers. September raw temperatures that match Octobers to the exact decimal place should be flagged at both GISS and NOAA. Anomalies 9C out of whack should be flagged.
These things can be correct, but they should be checked.
I should note that the ftp site I look at to find NOAA anomalies has not reported an October temperature. So, NOAA held their horses on processing the data and reporting. So did Hadley.
See
NOAA: ftp://ftp.ncdc.noaa.gov/pub/data/anomalies
Hadley: http://hadobs.metoffice.com/hadcrut3/diagnostics/global/nh+sh/monthly/monthly.land_and_ocean.90S.90N.df_1901-2000mean.dat .
As much as climate addicts like to see the processed data, it’s better for an agency to wait.
Of course, this issue will resolve itself. After all, no one really acts based on one month’s climate report. So, next month, we’ll know how far out of whack GISS’s report was. But, I don’t think it’s fair to just blame NOAA/NCDC. This sort of mistake is something programmers at GISS should be able to catch.
REPLY: The root of this problem lie is the antiquated multiple black box style of spaghetti code that is GISTEMP. And the fact that it is FORTRAN, rather than a modern language. Since the provenance of the code is owed to multiple programmers that have come and gone (Their model software is even worse) I think they’ve basically lost control of it. Hell when they released it last autumn, nobody could get it to compile and run. Big red flag.
I have a similar project for weather bulletins that was born in 1998. 10 years later it is unmanageable for the same reasons and we are starting over from scratch with a fresh project. GISS should have done the same long ago. – Anthony
If GISS’s temperature is affected (as it appears to be)
Well, I did notice one of their graphs looking down its nose at me.
This seems to be a major problem. Is there anybody out there who would have the time and resources to go over the “root” or station data for a longer period? Is it only sept-oct 2008? What other anomalies, like significant differences with WeatherUnderground, would be found? There’s enough work for a major research project!
Here in Canada, Nov 11th is a federal holiday (Remembrance Day). Is it also a holiday in the US? I took 1 day of vacation time Monday the 10th, to give myself a “4-day-weekend”. If a lot of people did something similar at NOAA and GISS, then maybe the people who should’ve caught the problem were not in. The computer output wasn’t halted, and it sailed through unchallenged.
NOAA normally puts out their Global dataset around the middle of the month. So the fact that they have not put anything yet (other than USA data) is not unusual. However, the GHCN error will propagate into their GMST as well.
Reply: The GHCN data as of Nov. 10 was generated from .dly files with November temperature records.
Wow! It seems that they are trying to hide the tracks and it might be hard to carry out the work I proposed in my posting at 08:34…
From Climate Audit commentary No.44:
PaulM:
November 11th, 2008 at 9:55 am
GISS has now deleted the duff station data.
But the incorrect 0.88 anomaly is still there in Fig.C.txt, as is the ‘red Russia’ graph.
I wonder whether they will acknowledge the error, or try to pretend that it never happened?
I was talking about this NOAA release:
http://www.ncdc.noaa.gov/oa/climate/research/2008/oct/global.html#introduction
which has the GHCN as a large part of the data.
One can’t generate any October maps either at http://data.giss.nasa.gov/gistemp/maps/
What I don’t get is this reliance on measuring devices spread throughout the countries, some of which are not known for even reasonable monitoring of such things) that are uncalibrated prior to each and every measure taken. When I was doing my research, I calibrated the electrical impedance of the electrodes prior to each and every measurement taken from the subjects. I also calibrated the signal I was using, again each and every time I used it. This step was recorded in the data base so that the final results could be verified as being reliable. Any change I measured with the electrodes could be said to be the result of the subject’s brainstem response and not the measuring device or calibration problems of the signal going in.
With surface stations, many have disappeared from the grid (potential bias). They are not calibrated on a daily basis (potential bias). Missing data is entered as a made up value instead of just “missing” (potential bias). The data is being handled by a group of people who want it to come out a certain way (potential bias). Am I missing something here? Has standard research design changed that much? How is it that this data even sees the light of day?
[…] Hansen and His Merry Band of NASA/NOAA Socialist “Scientists” continue busily packing the booster rockets for The Obamessiah’s Algore I economy-destroying ICBM (Intentionally-Caused Bolshevik […]
Fred, do you have a cached copy? NOAA has removed the online version.
[…] on in temperature reporting. The scientists are ‘laughably’ playing games with the data, and it shows. Or should that be ‘laughably’ called […]
NOAA has not released their October Global report yet per above link(which is not unusual as they normally do it the middle of the month). So there is no cached version available. Hopefully the folks at the NCDC will catch their error in the GHCN monthly database and NOAA will capture more realistic data in their report.
[…] Todas las mediciones oficiales apuntan a que 2008 será uno de los años más fríos de la última década. Sin embargo, el Instituto Goddard (GISS), perteneciente a la NASA, uno de los principales organismos de referencia del Panel Intergubernamental sobre el Cambio Climático de la ONU (IPCC) acaba de anunciar que el pasado mes de octubre ha sido uno de los más calurosos de la historia. En concreto, según sus mediciones, la temperatura global registró el pasado mes un aumento de 0,78 grados centígrados respecto a la media de referencia. Una anomalía muy significativa, pero que no encaja con la evolución histórica que viene registrando el aumento medio de la temperatura global en octubre a lo largo de los últimos años (por debajo de 0,6 grados centígrados, y descendiendo), tal y como recoge el portal WattsUpWithThat. […]
“And the fact that it is FORTRAN, rather than a modern language. Since the provenance of the code is owed to multiple programmers that have come and gone (Their model software is even worse) I think they’ve basically lost control of it.”
I believe, Anthony, from reading over at CA that there is even some more modern Python thrown in there for good measure. So it seems someone is hacking at it with some more modern tools but overall you seem to be correct in your conclusion that at this point it is just a mess.
I believe they finally did get something close to the posted source code to run, maybe after converting many of the routines to another language, I don’t remember now. But I don’t think they ever got the thing to spit out identical results to what Hansen gets.
One of the problems with the thing seems to be that missing values are “filled” using averages over time. That means that an anomalous reading today change the average and therefore values far in the past. That is probably why the huge anomaly this month bumped 2005 back up; it changed the average value that was used to fill in some missing data for that year. While that could effectively reduce the slope of warming by raising past temperatures when recent temperatures rise, the change would have differing impact depending on the amount of missing data in the past. Also, Hansen seems to set a “break point” of sorts for some stations where temperatures appear to get adjusted upwards after that point, and downwards before. That effectively increases or exaggerates the warming.
Anyway, GISS is, as you said, a mess.
Has this happened before? Is this a measure of how many stations currently on the grid that are unable to report because they do not work anymore? If this is a one-time problem, why are so many stations apparently not reporting this time around? Has the data been scrubbed prior to September for similar problems and this one just got through by mistake? Exactly how many current stations report 999 for any one month but we don’t see that till after it has been scrubbed? How many 999 months does a station have to go through before it is finally kicked off the grid? I remember the last time the data base was reduced. It didn’t happen gradually but more all at once. And temp averages changed because of it. I have a hunch that worldwide, there are as many stations not working anymore as there were back then. But is there a reason why they are keeping them on the grid this time instead of dumping them like they did earlier in the last century?
evanjones. You have my sense of humour – and I’d like it back please.
[REPLY – Sorry, sir. Your receipt is not in order. (On second thought, we could always just share it.)]
[…] many people will have read there was a glitch in the surface temperature record reporting for October. For many Russian […]