Guest Post by Willis Eschenbach
According to an article in the Hindustan Times by someone for whom English is a second language, I find:
Senior scientists at the Wadia Institute of Himalayan Geology (WITG) has rejected the Global Warming Theory and told that the Himalayas are quite safer zone on earth, where Global Warming has no role in controlling the conditions.
In an exclusive chat with HT, Director WIHG Dr AK Dubey has said that the conditions of Himalayas are controlled by the winter snowfall rather than external factors like much hyped Global Warming. He told that for a concrete result, at least 30 years of continuous research with steady outcome is needed to confirm the actual impact.
“According to a data for over 140 years available with a British weather observatory situated in Mukteswar (2311m) in Almora has actually revealed that temperature in that region witnessed a dip of .4 degrees,” he said.
So, as is my wont, I figured I’d go take a look. To distinguish urban from rural sites, GISS uses a “brightness index” which shows how much light comes from around the site as seen from a satellite. GISS lists Mukteshwar Ku as having a brightness index of zero, so they treat it as a rural station. Here’s the location per the GISS data, at 29.47°N, 79.65°E. It definitely appears to be a rural site.
Figure 1. Aerial View of the Mukteshwar Ku Surface Station locality.
Having seen the problems that occurred in Matanuska due to the application of a computer algorithm without quality control and checking, I next went to look at the record. Here is the GISS record for Mukteshwar Ku, before it has been subjected to the “homogeneity adjustment”:
Figure 2. GISS record of the temperature at Mukteshwar Ku before homogeneity adjustment
There’s a couple of oddities here. First, Dr. Dubey said that there were 140 years of temperature records from the station, but the GISS data covers 1897 to the present, or 113 years including the missing years.
In addition, it is clear that there has been some kind of serious change in the station. It is missing data from about 1993 to 1998, and when it starts up again the temperatures are much warmer than when it left off. (I can’t say exactly what years are missing, because curiously, the GISS server comes up with a “404 Not Found” when I ask it for the actual data.)
Seeing such an obvious problem with the data, I looked at the graph showing the temperature after homogenization to see how they had dealt with the problem … foolish me. I forgot that it was a rural station (brightness = 0), so it wasn’t adjusted at all. Sad to say, that’s the data that they used.
I’m used to not finding the data where I expect it to be, so to continue my analysis I just digitized the GISS graph so I could look at the effect of their leaving the data uncorrected. The gap was as I estimated, 1993-1998. Here’s that result:
Figure 3. Final GISS record of the temperature at Mukteshwar Ku. Note the difference in the trends when the recent data is included. Photo is of Nanda Devi Peak from Mukteshwar Ku.
As I said in my article about Matanuska cited above, the problem is that you can’t just devise a method for computer adjusting temperature data, apply it to all of the world’s stations, and call the job done. You need to look at and consider each and every station, as they are as individual as human beings. This is called “quality control”, and it is sadly lacking in all three of the major global temperature records (GISS, CRU, and GHCN).
Does this invalidate the GISS global temperature record? No. However, it does mean that they are not doing their job. They haven’t removed an obvious inconsistency in this case. How common is this type of problem? I don’t know.
But until they start over and do it right, it does mean that, like the baseball records of players who are known to have used steroids, the GISS global temperature has to be entered in the record books “with an asterisk” to indicate that lingering questions still remain.
********
Willis Eschenbach (23:42:49) :
Wren (23:18:18)
I don’t doubt errors in station data exist, but I imagine they are offsetting. If there was a systematic effort to bias surface records in favor of higher temperatures, GISS and UAH trends would diverge over the 1978-2009 period, but they don’t.
GISS trend 1978-2009 = 1.6°C/century
UAH trend 1978-2009 = 1.3°C/century
Difference, GISS – UAH = 0.3°C/century
GISS 20th century warming = 0.55°C/century
Since the difference between them is more than half the last century warming, I would hardly say that GISS and UAH “don’t diverge”.
*******
There’s more to it than even that — by standard GHG theory, mid-tropospheric sat temp trends should be greater than surface trends. Nearly twice as much.
Wren (01:20:39)
…
Well, I don’t know exactly what you did to come up with those numbers, but regardless, I agree you can make a small difference look bigger if you enlarge it.
What I “did to come up with those numbers” is called mathematics.
I went and got the GISS and the UAH data. Click the links, you can download the data yourself. I calculated the per-century trends in each one. I did not “enlarge” anything as you fatuously claim. You can do the math yourself … or perhaps not, I don’t know.
But that, as Jack Webb used to say, is “Just the facts, ma’am”. The difference in the GISS and UAH trends is about a third of a degree per century. Don’t like it? Sorry … math is either right or wrong, it doesn’t depend on your dislikes. If you can show my math is wrong, go for it. If not, live with it.
w.
========
You enlarged the difference between temperature data for UAH and GISS by calculating a per century trend. A century of actual comparability for the two does not exist. Comparable data only go back to Dec. 1978, which is when UAH began. You simply do not know what UAH data would show if it were available for a century.
Besides, your results would be different if you compared the most recent month with the beginning month. Here are the numbers from the tables you cited, with UAH in the first column and GISS in the second :
Dec. 1978 -0.20 -4
Jan. 2010 .63 71
The January 2010 UAH number is a recent revision, but I do not know whether the December 1978 number has also been revised, so the difference may be greater than -0.20 to 0.63.
climategatestuff (03:55:04)
My apologies for the lack of clarity. What I meant was that he personally has that amount to spend on his projects, not that he is putting it in his pocket. However, I have also read that there’s only 3 salaried folks at CRU, so a good chuck of grant monies is going into salaries.
Nick,
The business model you describe is very similar to the one I work under. Our company is an engineering services company. Our output is our labor. We get paid to solve problems. Now, if one of our salaried people brings in large amounts of money to the organization then the company will inflate the salary of the “triple threat” because the cost of the increase can be spread out over the entire project. If one keeps making the big kill, then one is rewarded accordingly.
Willis,
In the U.S. the NCDC made an algorithm to identify step-change inhomogeneities like the one identified here and automatically correct. The application of this method to USHCN v2 actually increased temp adjustments a fair bit vis-a-vis USHCN v1.
Now that a serious effort to produce GHCN v3 is underway, its my understanding that they plan on producing a v3_adjusted version of the data that has a step-change inhomogeneity correction applied. It should be interesting to see how it turns out.
Also, I’m sure the NCDC folks would agree with your wholeheartedly that GHCN station metadata could be much improved, and that more folks should be funded to work on it.
REPLY: Much improved? There’s hardly any GHCN metadata beyond a sometimes truncated place name and coarse lat/lon. NCDC didn’t even start adding international stations to the MMS database until late 2007. – A
beng (10:16:45)
Actually, the situation is quite complex. See my analysis of the subject here on WUWT.
Wren (01:20:39)
So your claim is that a trend of one degree per century is larger than a trend of one tenth of a degree per decade …
Wren, you are free to say that a trend of .13°C/decade (UAH) and .16°C/decade (GISS) don’t diverge.
I say that they diverge by 0.03°C per decade, or 0.003°C per year, or 8.3E-5°C per day, or 0.3°C per century.
None of those is different from the others. None is “enlarged”. None of them depend on how long the dataset is.
Wren (10:18:53)
Wren, comparing the first and last months of a dataset to determine the trend is generally considered A Bad Thing™ by mathematicians. This is because it is too sensitive to the endpoints. Suppose, for example, that a trendless (no warming or cooling) fifty year temperature dataset starts in December and ends in July. If we compare the first and last months, we’d think that there was a huge trend, because July is much warmer than December.
My 2 cts about Mukhteshwar – its position:
A Thai site shows other location, which looks quite acceptable:
http://thai.wunderground.com/global/stations/42147.html?MR=1
Gladstone Family web does not consider the database coordinates correct:
http://weather.gladstonefamily.net/site/42147
OT but this seriously sucks . . .
http://www.telegraph.co.uk/science/science-news/7414334/Large-Hadron-Collider-to-close-for-a-year-for-refit-and-repairs.html
Willis,
Dont trust the metadata, yet
http://rhinohide.wordpress.com/2010/03/10/ghcn-metadata-horseshoes-and-hand-grenades/#comment-101
@ur momisugly thethinkingman (11:52:04) :
Not unexpected for the LHC, given its problems (see here). It is an immense highly complex device with many parts that have to precisely work together in a synchronized fashion for it to work at all, with an operating limit as designed that was really pushing the upper limits of the components to begin with. It will be a miraculous event if it ever functions as powerfully as it was promoted to be.
Speaking of such, how is the NIF ICF project coming along? 😉
http://www.unur.com/climate/ghcn-v2/207/42147-zoomed.png
Here’s the non-annual graph, in which it appears the “warming” is nothing more than slightly milder winters, but no increase in summer temps. One should look at local population changes and factor in Spencers UHI correction…
Note: The locals are already adjusting the data, note the black and red lines nearly overlapping in the graph, which they admit to on their website:
http://www.unur.com/climate/ghcn-v2/207/42147.html
Now, look at the raw monthly data for the period in question:
http://data.giss.nasa.gov/work/gistemp/STATIONS//tmp.207421470003.0.1/station.txt
THE MET-ANN data shows COOLING, NOT WARMING. The GISS adjusted data is seriously manipulated.
As it is grant, not salary, there is no tax other than when it is spent (as in sales, VAT and other taxes). However, it is HIS to spend as he sees fit. In other words, what he has been paying out of pocket for in the past (meals, transportation, over night stays) can now be charged against the grant. It makes for some mighty fine dining!
When I was traveling for work, I was given a meal allowance that was very generous. Needless to say, I did eat better than when I was dining at home.
steven mosher (12:01:30)
Interesting, Mosh. I was particularly intrigued by the statement that the change to brightness affected a quarter of the stations. I hadn’t considered the effect that the mislocation of the stations would have on the brightness. I have seen before that the coordinates sometimes end the station up in the ocean (brightness = 0, presumably), so this is not a trivial problem.
Again, it highlights the idiocy of not doing station-by-station quality control. They seem to just grab an unchecked database, apply an unverified computer program, and report the results without any QC. No checking at any of the three stages of the process.
It’s the old story of GIGO made new … now it stands for “Garbage In, Grants Out” …
Willis Eschenbach (11:05:40) :
Wren (01:20:39)
You enlarged the difference between temperature data for UAH and GISS by calculating a per century trend. A century of actual comparability for the two does not exist. Comparable data only go back to Dec. 1978, which is when UAH began. You simply do not know what UAH data would show if it were available for a century
So your claim is that a trend of one degree per century is larger than a trend of one tenth of a degree per decade …
Wren, you are free to say that a trend of .13°C/decade (UAH) and .16°C/decade (GISS) don’t diverge.
I say that they diverge by 0.03°C per decade, or 0.003°C per year, or 8.3E-5°C per day, or 0.3°C per century.
None of those is different from the others. None is “enlarged”. None of them depend on how long the dataset is.
—
My claim is exactly what I said. You simply do not know what UAH data would show if it were available for a century.
mikelorrey (13:27:28)
Good find, Mike. What do the Indians give as the reason for the adjustment?
(As an aside, the GISS link is a temporary link. You can get the data by going here. Make sure you select the type of dataset first, individual, combined, or homogenized.)
I’m not sure what you are calling the “GISS adjusted data”. There are four overlapping records covering 1897-1993. Here are their trends, using the MetAnn (meteorological year, which is December to November).
Record 1: 1897-1991, 0.016°/decade
Record 2: 1949-1990, -0.16°/decade
Record 3: 1971-1980, 0.014°/decade
Record 4: 1987-2010, (missing 1993-1998)
When I combine records I use the “first differences” method. My combination of the records gives the same trend as the GISS combined data.
Here’s the strange part, though. In the first part of the time period, where there is only one record (1897-1948), the GISS combined record agrees completely with the GISS record 1. That would be expected. All of the numbers agree perfectly, the monthly numbers, the numbers for the four seasons … all except the numbers for the MetAnn average. Of the 53 years, fifrteen of them differ between the two records.
The differences are small, one or two hundredths … but I’m among other things an accountant. Stuff like that makes me nervous, again it indicates that they haven’t checked their results. It also makes the trend very slightly warmer. Of the 15 results, 13 are warming, 2 are cooling.
Finally, as you can probably guess, none of these trends are statistically different from zero.
Wren (10:18:53)
Besides, your results would be different if you compared the most recent month with the beginning month. Here are the numbers from the tables you cited, with UAH in the first column and GISS in the second :
Dec. 1978 -0.20 -4
Jan. 2010 .63 71
Wren, comparing the first and last months of a dataset to determine the trend is generally considered A Bad Thing™ by mathematicians. This is because it is too sensitive to the endpoints. Suppose, for example, that a trendless (no warming or cooling) fifty year temperature dataset starts in December and ends in July. If we compare the first and last months, we’d think that there was a huge trend, because July is much warmer than December.
=====
Oh, I know comparing end points isn’t necessarily the best way to discern a trend. But neither is your way.
Your example is isn’t a good one, since the comparison I made started in December and ended in January.
The Indian plate colliding with the Asian plate means the himalayas are constantly being uplifted even as weathering occurs. The homogeneity adjustment has probably been applied as the weather station rises.
/sarcasm
The CRU, NOAA and GISS were formed to measure global warming. If there is no global warming then they cease to have a purpose, their functions could be absorbed elsewhere. Unless they keep on finding global warming they are out of a job.
Necessity is the mother of ‘invention’!
REPLY: CRU and GISS (in their current mission) fit that description, NOAA does not. -A
Graph of monthly GHCN data (as of 2007) for Mukteshwar Ku:
http://www.unur.com/climate/ghcn-v2/207/42147.html
Willis,
I was using the fourth period, the most recent data, which included the gap period. The post-gap MetAnn data doesn’t match whats charted, its actually cooler than the pre-gap data…. NM I was looking at it upside down…
Sinan,
I found your site earlier, thank you for providing this. Can you explain the gap in data in the early 90’s and if there is any change in the station siting?
This station is NOT rural. Here’s the google maps location, zoom in:
http://maps.google.com/maps?f=q&source=s_q&hl=en&geocode=&q=29.47N+79.65+E&sll=29.462318,70.65033&sspn=0.408328,0.856934&g=29.47N+70.65+E&ie=UTF8&ll=29.47,79.65&spn=0.00319,0.006695&t=h&z=18&output=embed&w=425&h=350
The population in the local area is significant and esp in recent decades should exhibit significant UHI here.
Willis:
My contact at the Wadia Institute for Himalayan Geology indicates that (a) they had to buy their data; (b) they could not share it; and, (c) metadata for sites was never available freely on the (IMD) website. I have not yet heard back from IMD.