Guest Post by Willis Eschenbach
Recapping the story begun at WUWT here and continued at WUWT here, data from the temperature station Darwin Zero in northern Australia was found to be radically adjusted and showing huge warming (red line, adjusted temperature) compared to the unadjusted data (blue line). The unadjusted data showed that Darwin Zero was actually cooling over the period of the record. Here is the adjustment to Darwin Zero:
Figure 1. The GHCN adjustments to the Darwin Zero temperature record.
Many people have written in with questions about my analysis. I thank everyone for their interest. I’m answering them as fast as I can. I cannot answer them all, so I am trying to pick the relevant ones. This post is to answer a few.
• First, there has been some confusion about the data. I am using solely GHCN numbers and methods. They will not match the GISS or the CRU or the HadCRUT numbers.
• Next, some people have said that these are not separate temperature stations. However, GHCN adjusts them and uses them as separate temperature stations, so you’ll have to take that question up with GHCN.
• Next, a number of people have claimed that the reason for the Darwin adjustment was that it is simply the result of the standard homogenization done by GHCN based on comparison with other neighboring station records. This homogenization procedure is described here (PDF).
While it sounds plausible that Darwin was adjusted as the GHCN claims, if that were the case the GHCN algorithm would have adjusted all five of the Darwin records in the same way. Instead they have adjusted them differently (see below). This argues strongly that they were not done by the listed GHCN homogenization process. Any process that changed one of them would change all of them in the same way, as they are nearly identical.
• Next, there are no “neighboring records” for a number of the Darwin adjustments simply because in the early part of the century there were no suitable neighboring stations. It’s not enough to have a random reference station somewhere a thousand km away from Darwin in the middle of the desert. You can’t adjust Darwin based on that. The GHCN homogenization method requires five well correlated neighboring “reference stations” to work.
From the reference cited above:
“In creating each year’s first difference reference series, we used the five most highly correlated neighboring stations that had enough data to accurately model the candidate station.”
and “Also, not all stations could be adjusted. Remote stations for which we could not produce an adequate reference series (the correlation between first-difference station time series and its reference time series must be 0.80 or greater) were not adjusted.”
As I mentioned in my original article, the hard part is not to find five neighboring stations, particularly if you consider a station 1,500 km away as “neighboring”. The hard part is to find similar stations within that distance. We need those stations whose first difference has an 0.80 correlation with the Darwin station first difference.
(A “first difference” is a list of the changes from year to year of the data. For example, if the data is “31, 32, 33, 35, 34”, the first differences are “1, 1, 2, -1”. It is often useful to examine first differences rather than the actual data. See Peterson (PDF) for a discussion of the use of the “first-difference method” in climate science.)
Accordingly, I’ve been looking at the candidate stations. For the 1920 adjustment we need stations starting in 1915 or earlier. Here are all of the candidate stations within 1,500 km of Darwin that start in 1915 or before, along with the correlation of their first difference with the Darwin first difference:
WYNDHAM_(WYNDHAM_PORT) = -0.14
DERBY = -0.10
BURKETOWN = -0.40
CAMOOWEAL = -0.21
NORMANTON = 0.35
DONORS_HILL = 0.35
MT_ISA_AIRPORT = -0.20
ALICE_SPRINGS = 0.06
COEN_(POST_OFFICE) = -0.01
CROYDON = -0.23
CLONCURRY = -0.2
MUSGRAVE_STATION = -0.43
FAIRVIEW = -0.29
As you can see, not one of them is even remotely like Darwin. None of them are adequate for inclusion in a “first-difference reference time series” according to the GHCN. The Economist excoriated me for not including Wyndham in the “neighboring stations” (I had overlooked it in the list). However, the problem is that even if we include Wyndham, Derby, and every other station out to 1,500 km, we still don’t have a single station with a high enough correlation to use the GHCN method for the 1920 adjustment.
Now I suppose you could argue that you can adjust 1920 Darwin records based on stations 2,000 km away, but even 1,500 km seems too far away to do a reliable job. So while it is theoretically possible that the GHCN described method was used on Darwin, you’ll be a long, long ways from Darwin before you find your five candidates.
• Next, the GHCN does use a good method to detect inhomogeneities. Here’s their description of their method.
To look for such a change point, a simple linear regression was fitted to the part of the difference series before the year being tested and another after the year being tested. This test is repeated for all years of the time series (with a minimum of 5 yr in each section), and the year with the lowest residual sum of the squares was considered the year with a potential discontinuity.
This is a valid method, so I applied it to the Darwin data itself. Here’s that result:
Figure 2. Possible inhomogeneities in the Darwin Zero record, as indicated by the GHCN algorithm.
As you can see by the upper thin red line, the method indicates a possible discontinuity centered at 1939. However, once that discontinuity is removed, the rest of the record does not indicate any discontinuity (thick red line). By contrast, the GHCN adjusted data (see Fig. 1 above) do not find any discontinuity in 1941. Instead, they claim that there are discontinuities around 1920, 1930, 1950, 1960, and 1980 … doubtful.
• Finally, the main recurring question is, why do I think the adjustments were made manually rather than by the procedure described by the GHCN? There are a number of totally independent lines of evidence that all lead to my conclusion:
1. It is highly improbability that a station would suddenly start warming at 6 C per century for fifty years, no matter what legitimate adjustment method were used (see Fig. 1).
2. There are no neighboring stations that are sufficiently similar to the Darwin station to be used in the listed GHCN homogenization procedure (see above).
3. The Darwin Zero raw data does not contain visible inhomogeneities (as determined by the GHCN’s own algorithm) other than the 1936-1941 drop (see Fig. 2).
4. There are a number of adjustments to individual years. The listed GHCN method does not make individual year adjustments (see Fig. 1).
5. The “Before” and “After” pictures of the adjustment don’t make any sense at all. Here are those pictures:
Figure 3. Darwin station data before and after GHCN adjustments. Upper panel shows unadjusted Darwin data, lower panel shows the same data after adjustments.
Before the adjustments we had the station Darwin Zero (blue line line with diamonds), along with four other nearby temperature records from Darwin. They all agreed with each other quite closely. Hardly a whisper of dissent among them, only small differences.
While GHCN were making the adjustment, two stations (Unadj 3 and 4, green and purple) vanished. I don’t know why. GHCN says they don’t use records under 20 years in length, which applies to Darwin 4, but Darwin 3 is twenty years in length. In any case, after removing those two series, the remaining three temperature records were then adjusted into submission.
In the “after” picture, Darwin Zero looks like it was adjusted with Sildenafil. Darwin 2 gets bent down almost to match Darwin Zero. Strangely, Darwin 1 is mostly untouched. It loses the low 1967 temperature, which seems odd, and the central section is moved up a little.
Call me crazy, but from where I stand, that looks like an un-adjustment of the data. They take five very similar datasets, throw two away, wrench the remainder apart, and then average them to get back to the “adjusted” value? Seems to me you’d be better off picking any one of the originals, because they all agree with each other.
The reason you adjust is because records don’t agree, not to make them disagree. And in particular, if you apply an adjustment algorithm to nearly identical datasets, the results should be nearly identical as well.
So that’s why I don’t believe the Darwin records were adjusted in the way that GHCN claims. I’m happy to be proven wrong, and I hope that someone from the GHCN shows up to post whatever method that they actually used, the method that could produce such an unusual result.
Until someone can point out that mystery method, however, I maintain that the Darwin Zero record was adjusted manually, and that it is not a coincidence that it shows (highly improbable) warming.
Sponsored IT training links:
Want to pass HP0-J33 at first try? Gets certified 000-377 study material including 199-01 dumps to pass real exam on time.
Nick Stokes
RE my post Geoff Sherrington (22:40:07) : on 20/12
This refers to a 7-year BOM study which BOM documentation shows that the old Darwin site and the Airport location have identical mean temperatures. I have said that there could be reporting errors, though I suspect none will make a difference, for the BOM continues to make the comparison available.
If you have access to more information than I do (I have very little – I have to search for it), then perhaps you can explain why it is a good idea to put a step change into the data at about the 1940 changeover.
Ooooh, new information. I love new information.
I just paid the outrageous sum of $29.95 for Peterson’s paper describing the GHCN homogenization method. This is CREATION OF HOMOGENEOUS COMPOSITE CLIMATOLOGICAL REFERENCE SERIES, THOMAS C. PETERSON AND DAVID R. EASTERLING, INTERNATIONAL JOURNAL OF CLIMATOLOGY, VOL. 14. 671-679 (1994).
Regarding reference series and distance from the station in question, the paper says:
Now, for Darwin, the hundred nearest stations includes stations out to 1,656 km from Darwin. Since I had already looked at the stations out to 1,500 km from Darwin, this did not add many possibilities. Here are the first difference calculations for the final few of the possible candidate stations per Peterson for the 1920 adjustment to Darwin:
PALMERVILLE = -0.39
BOULIA = 0.31
NULLAGINE (NULLAGINE POST OFFICE) = -0.01
COOKTOWN AMO = -0.22
PORT DOUGLAS POST OFFICE = -0.07
LOW ISLES = -0.35
HERBERTON POST OFFICE = -0.31
As you can see, none of these stations are suitable. In addition to having low correlations, the correlation in all of these is negative, and Peterson says stations with negative correlation can’t be used.
So that complete the demonstration. We cannot construct a reference series for the Darwin 1920 adjustment using the GHCN method and criteria. There’s only three candidate stations with positive correlations for that time, and they have a correlation of 0.06, 0.35, and 0.35 … far too low for a reference series which, according to Peterson, must have a correlation of 0.80 with Darwin.
JJ, you happy now?
JJ:
GHCN does not appear to have a numeric limit on the distance for ‘neighboring’
JJ:
GHCN can apprently use any station within GHCN’s definition of ‘region’
I think, I got the point: “Neighboring” and “region” for GHCN does mean: somewhere in the same climatic region.
Ok, let it be so. But this brings us immediatedly to another severe problem. As I understand it the GHCN method is the following: 6 stations in this “region2 are behaving the same way, i.e. being correlated >0,8. The data of 5 are used to “correct” the data of No. 6. In the case of thermometer faults, station movements, all is fine, and this method can give hints to explore the chronic of No 6 for such effects. For Darwin is this for the year 1939 or 1940.
But if the deviation at No. 6 is caused by different reaction on climatic changes at its specific location, the latter will be wiped out! Then we end up which a situation we call in German: Die Katze beisst sich in den Schwanz (The cat bites in its own tail; I dont know an appropriate english expression). We start with the definition of “climatic region”, but if a station shows deviant behaviour because of different local climate, it will be corrected and corrected and corrected again, until it follows the proposition of being in the same “region”.
Best regards from Germany and Austria (not Australia ;-))
Geoff Sherrington (20:16:46) :
Nick Stokes,
Your response to my observation that the BOM declines to use pre-1950s data for Coonanbarabran is?
None. I don’t know who’s right. All I’m pointing out is that the mechanics that produce the Darwin rise can also produce the Coona fall. And these are outliers. They are in the outer 1% or so.
Dave F (19:38:26) :
why adjust the data in the first place?
Good question. GHCN themselves say that for a lot of purposes, you might prefer the unadjusted data. They say the adjustments help with regional attribution.
But put it the other way around. What would you folk be saying if they made changes, updated equipment etc, and didn’t adjust.
3×2 (05:46:54) :
I’m still left with a couple of questions.
For those who say that individual stations don’t matter, the global stats will sort things out
If, when scaled globally, local “adjustments” have no real impact … why bother “adjusting” in the first place? Adjustments at the individual station level obviously do impact both regional and global results. Especially if the end result is in the 1/10th degree range.
Specifically for Darwin adjustments.
If you take the difference between the “winter” and “summer” means over the entire record (25/28 from v2.mean) as being about 3°C. Is a 2.25°C “adjustment”, in any direction, at any point, even physically possible?
I would expect so, in ’41 the station was moved from the old PO (destroyed by bombing shortly after) to the airport. In the dry season the prevailing wind at Darwin is southeasterly, the old PO was situated on the end of a peninsula so the southeasterlies blow over water onto the land. This is shown by the data, the record low at the airport is 10.4, the record at the PO was 13.4. The move was made because the PO site was no longer satisfactory due to overhanging trees, subsequent to ’41 the site was relocated several times on the airport.
Nick Stokes (23:24:59) :
Nick, I don’t know about Dave, but I have no problem with adjusting for known discontinuities. Such things as equipment changes and station moves definitely introduce false signals into the data.
However, the devil is in the details. My own feeling is that we should take the first difference of the data, and eliminate the one observation from the year of the discontinuity. That way, the trend of the data (which is all that we are interested in) is maintained, and we can average the first difference of the observations directly with other stations to get an average trend over a period of time.
The averaging, of course, introduces other difficulties … but that’s a subject for another thread. My preferred method also doesn’t deal with slow changes, like buildings and trees and paving around the station … but then the GHCN method doesn’t deal with those either.
w.
GHCN have to adjust the data because they have almost no sites that have not changed over the last 100 years. They have sites that have had new screens, they have sites that went from mercury to electronic thermometers, they have sites that were in small villages 100 years ago that find themselves in large cities today, they have sites that had horse and cart going by in 1900 and then three lanes of articulated trucks in 1980, they have sites that were near coastal lighthouses that moved to being near inland airports. All sorts of changes that affected 99% of all the sites. What is worse, from a GHCN point of view, is that the few sites that didn’t change much over 100 years show no warming at all! They should be held up as “gold standards” but GHCN prefers to sublimate them under a mountain of dodgy data from less reliable sites.
Now, if GHCN could reliably detect when the site changes occured, they could simply look for the trend in temperature between those changes when it can be assumed that no changes to the measurement of any kind occurred and therefore the relative temperature change over time could be considered accurate. The fact that they choose not to adopt this simple approach speaks volumes in itself – copying data from one site 1500km away to another site has no scientific validity at all, especially when it intoduces huge adjustments in the trend to the final output. Remember, it is the trend that actually matters, not the absolute temperature. We know today’s absolute temperature and that’s all we need to know if we know the trend, so an adjustment to an absolute temperature that makes a dramatic change in the observed trend is a very bad adjustment approach indeed.
UHI must also be considered, but this can be detected at each site by direct measurement within the UHI affected area and just outside. If the UHI is measured this way over one year then it canb be used to introduce a downard uncertainty in any measured warming trend.
Here is how this particular aspect of the fraud works. Lets say we have two sites that in 1900 were edge of town. By 1960 these towns have expanded and there is heavy traffic causing the urban heat island effect. We then decide that a site is so bad that we move one site out of this urban heat island effect, and the other remains where it is, thus affected by the urban heat island effect.
The algorithm used here is looking for discontinuities, so it will detect the discontinuity in the site that is moved. It will detect no discontinuity in the site that was not moved, but which is affected by urban heat island effect. This is because the urban heat island effect will grow gradually over time and will be indistinguishable from other climate effects.
The algorithm will detect that the site that has been moved has dubious data. It will then seek to correct the data with nearby sites that have not shown a discontinuity. If this is the site with the urban heat island effect, it will copy the data from that site to the moved site and smooth it into one continuous trend.
The impact of this is that rural sites will appear to show a warming trend (which is asctually due to the urban heat island effect of neighbouring sites). Since the site is now rural it will appear as if it can’t be affected by urban heat island effect, but in fact it will have been affected by the copying of data from precisely those sites that are affected by urban heat island effect.
This algorithm is biased in favour of trashing rural site data unaffected by urban heat island effect in favour of data that is affected by urban heat island effect. It has no credibility whatever.
Nick Stokes (17:36:22) :
“There’s nothing to indicate that your frost hollow story involves the Coona weather station. The town is by the Castlereagh river, but in a fairly flat region.”
Coona is a cold hole in winter. It’s surrounded by a ring of low hills apart from the path of the Castlereagh. I don’t doubt at all that it’s in a frost hollow, although the magnitude of the cooling surprised me. It also has the Warrumbungles immediately to the west.
You might be confusing it with Coonamble which is also on the Castlereagh and *is* dead flat. It’s only about 100km NW of Coona, so it’s easy to confuse them if you aren’t familiar with the area.
Nick,
Your point that Darwin is among a small percentage of outliers may be pointless in light of the fact that not all stations are equal in the impact they have on the global average. As it stands now Darwin is a high impact station where the adjusted data contains a very large and very false warming trend inserted.
If you could provide an example of a similarly high impact station where the adjusted data has a false cooling trend of similar magnitude (-0.6C/decade) then your point about the flaws in the adjusted data cancelling each other out** might be valid. Otherwise that dog doesn’t hunt.
**They still don’t cancel out completely. It’s already been accepted at face value that the adjustments result in an average bias of approximately 0.2C/century of additional warming. That is a significant number when we’re trying to dig out what, if any, human-caused warming has happened.
So far you’ve persuaded me that at least 0.2C of the last century’s warming isn’t man-made but rather is Mann-made, if you know what I mean, and I think you do.
Ryan Stephenson (02:45:27) :
I certainly agree with your approach.
The idea of using stations up to 1500 km or more away “to adjust” a site seems very “dodgy” at best. I do not know Australia, but the whole idea that weather in one location is similar to that in another does not pass the smell test. Within 100 miles of where I live we have oceans, lakes, the piedmont and sand hills. Five miles away ALWAYS gets more rain that I or the airport weather station down the street does. The idea might work OK on a flat plain but not on mountains and ocean front areas. And yes I do understand a significance test for correlations. As a test to determine if the station may have a problem that needs investigation – fine. As the basis for “adjustments” NO.
The whole thing reeks of lazy data handling techniques.
Nick Stokes (23:24:59) :
I am aware of that, my point was not that the data shouldn’t be adjusted, just that you are saying that it does not contribute much. If that were true, then wouldn’t there be a minute difference in the way unadjusted and adjusted data look plotted next to each other for the entire GHCN dataset? Is there a .0175C difference in the trend when you do that, which is what you and GG are saying in your histogram?
Willis Eschenbach (22:57:54) :
“”our only test based solely on distance was limiting the list of potential reference stations to the 100 stations nearest to each candidate station.””
Great work, Willis.
1. I think you’ve shown that the Peterson algorithm couldn’t have been followed. (Can you just tell us where you read about the 0.8 correlation requirement? Isn’t it possible that they accepted a correlation of 0.4 as long as it was the “highest correlated neighboring station?”)
2. I think you’ve also shown at least one flaw in the Peterson algorithm even if it is utilized objectively. Testing 100 stations for adequate correlation doesn’t seem reasonable. As you hinted at before, even if 1,000 adjustments are using correlations with a p-value = 0.05, then that implies 50 of those 1,000 correlations were a coincidence.
Overall, I think JJ did you a favor. He was definitely condescending, but he did push you to strengthen your argument quite a bit. However, I agree that it would be nice for him to help out with some of these efforts. He’s obviously smart enough. Unless he’s working on something of his own right now. JJ, do you care to share?
Nick Stokes, I appreciate some of the analysis you’ve done, but it seems that your entire argument boils down to the fact that Coonanbarabran is jumpier than Darwin.
All your histograms of trends are quite useless without being able to prove that Coonanbarabran was also manually adjusted. Sorry, but this can’t be done using a visual jumpiness test. Maybe we should start by looking at the nearest 100 stations and finding the most highly correlated.
@Willis: Have you tried taking the approach you/we suggest? Take the data between the discontinuities detected by the algorithm, then check the gradient of that data? Maybe it is difficult because there are only 10 or so datapoints and it is dependent on where you start and stop exactly, but it seems that the gradients between the discontinuities are about 0Celsius/Century compared to the gradient imposed by the corrections. Excel can plot a line of regression between the points, e.g. for the last 12 years we seem to have stable data and Excel can plot a line of regression between the last 12 points of the raw data and the gradient of this line can be derived. Sorry to dump the extra work on you but if you have the data tabulated already it should take you long.
That would really blow this whole thing out of the water. After all, what the algorithm itself is telling us is the site changed about three times after 1940 but was stable as a measuring site between those changes. So that is where the quality data is – the raw data between the site changes. Measure the gradient between the changes and its job done. It sure isn’t 6Celsius/century. So the algorithm will be shown to fail the sanity check badly.
Without having looked at this particular station specifically, a couple of points, just because A and B look similar doesn’t mean they will pass/fail the same statistical test.
Given an artificial cut off (e.g. p-value 0.01), it is possible that one station will “fail” the test (e.g p-value = 0.0098) and needs correction, and one that “looks” it like won’t (e.g. p-value = 0.011 (which means no correction).
The samething happens with the correlations. Even assuming that two stations that look a like both “fail”, there is no guarantee that they won’t be corrected differently. A and B are highly correlated. A is correlated to C with a value of 0.81. C is used to correct A. B has a correlation to C of a value of 0.79. C isn’t used to correct B. As C is likely highly correlated to whatever else is used to correct B, it is likely the correction is similar, but over a number of years with enough corrections, it is possible to start with two things that look very similar, but don’t end up being very similar.
The last issue is your given correlations. I’m not sure why 1915 is relevant, or giving the correlation over the whole rest of the life of the stations (e.g 1915-until the data ends) (which is what I presume you did as you didn’t give any years).
My reading of the work was that they were doing a “local” (in time) tests to do a “local” correction based on pretty “local” correlations (I don’t remeber the exact times frames, but 5 years sticks in my mind). In other words, if I’m going to correct something in 1945 (just an example), then I don’t need the correlation back to 1915. I do need a station that has a rather long data set (25 years sticks in my mind), but that I don’t think the that 25 years is even post-correction data (e.g. I could use a station that ran from 1940 to 1965 to correct an issue with another station with an issue starting in 1945).
There are issues with exactly reproducing what they did. For example, it wasn’t clear to me if they were doing correlations to find “good” stations based on the corrected data for the station or the uncorrected data for the station.
And you are right, they’ve AT BEST poorly defined “neighborhood” and what stations they weren’t able to find a “neighborhood” for and didn’t correct and which ones they did are unclear.
I generally don’t think they expected anybody to try and EXACTLY reproduce what they did when they published the method (the publication before this really exploded into the issue it has). They assumed that people would either simply take their corrected data and use it, OR take the raw data and come up with their own method.
I will point out that isn’t exactly rare. In biology, for example, papers rarely truely give all the details. They rarely state, for example, how solutions are stored, and what quality of water was used to make them.
I understand what you are trying to do, but it would be much more convincing if you developed a method based on the raw data that didn’t show warming. There were always be points of contention with any method (e.g. why use the time frame the did to do the tests, why use the p-value cut off they used to say a particluar set of data “failed” and need a correction, why use a correlation of 0.8, etc.).
If for a some set of stations, their method “fails” that isnt’ shocking and doesn’t change the total result.
A reasonable method that took the same data and showed that golbally warming isn’t happening would be a big deal.
Using sites 1500KM away?
This shows the difference in sites across the UK,
Ten years with the hottest months from the Met Office Station Data.
Order of maximum temperature
Oxford back to 1853
2006 1983 1995 1911 1921 1976 1868 1947 2003 1874
Stornaway back to 1873
1899 1947 1997 1880 1955 2006 1933 1955 1976 1901
Durham back to 1880
2006 1995 1887 1947 1983 1975 1989 1901 1940 1990
Sheffield back to 1883
2006 1975 1983 1995 1976 1947 1921 1934 1995 1887
Eskdalemuir back to 1911
2006 1947 1955 1983 1989 1995 1975 1976 1934 1940
temp (08:34:27) :
“”If for a some set of stations, their method “fails” that isnt’ shocking and doesn’t change the total result.””
I don’t think Willis was trying to show that their method failed. I think he was trying to show that they failed to use the method.
Yeah, but he hasn’t done that. One of his simple claims is that since one was corrected and the other wasn’t, and they look the same that is evidence that they didn’t use their method, but that isn’t true.
You actually have to run the method and see which if either fail the test that is described. If they all fail the test (again, fail meaning they need correction), then you have issues, and if none of them fail the test, then you have issues.
Once you’ve determine if (and where) they fail the test, then you can start to worry about identifying the relevent stations to do the correction.
To simply state that they didn’t because it doesn’t “look” they did, and without actually trying to carry out the steps they do describe doesn’t mean you’ve reached a valid conclusion.
temp (11:57:47) :
“”To simply state that they didn’t because it doesn’t “look” they did, and without actually trying to carry out the steps they do describe doesn’t mean you’ve reached a valid conclusion.””
Check is post at (22:57:54).
I think it completes the effort of trying to carry out the steps they do describe.
He tested the 100 closest stations (which is prescribed by the Peterson method) and found that none had a correlation adequate enough to make adjustments.
What more do you want him to do to prove that the method wasn’t objectively used?
Sorry, I meant, check HIS post at (22:57:54).
As I said originally, I am assuming he is giving correlations over the entire time series because he isn’t giving starting and ending years for producing the correlation, which isn’t what they did, I think.
If that is wrong, then fine, but it isn’t clear from what he has written.
temp (12:33:00) :
“”which isn’t what they did, I think.””
What do you think they did? Do you think they truncated time series in order to improve the correlation? If so, then many people agree with you, and you might be right.
So what exactly would you recommend? When calculating correlations of 100 neighboring stations to be used to test for a discontinuity in a specific year you will be required to perform thousands of correlation calculations by truncating each time series by one data point at a time.
Then, you’ll need to repeat that process in order to test for a discontinuity for each point of the station’s time series. 70 years worth of time series data discontinuity tests represents hundreds of thousands of calculations in order to test a single station record.
If you have a better solution for replication, then please share it.
Willis,
With vjones’s help and with the aid of EMSmith’s excellent documentation, I’ve been carrying out my own analysis of the NOAA GHCN data. My first step was to reproduce you excellent analysis for Darwin (unlike the Team who think that ‘there’s nothing to see here, move on’). I’ve therefore been applying the scientiic method and have attempted to falsify your analysis. I’m sorry (actually I’m glad) to say that I failed! I’ve reproduced your charts and results almost 100% and have documented my efforts on vjones blog ‘diggingintheclay‘. You can read the thread in which I reproduce your analysis by clicking on the link below.
Reproducing Willis Eschenbach’s WUWT Darwin analysis
As I’m sure you already know and appreciate science progresses by ‘standing on the shoulders of giants’ so I’ve taken the liberty of further extending you excellent analysis for Darwin to all the WMO stations in the NOAA GHCN dataset.
Specifically I’ve attempte dto answer the question posed by others on your original Darwin thread as to whether or not Darwin is a special case or not?
Well judge for yourself by clicking on the link below which documents my extension of your analysis to include the whole NOAA GHCN dataset.
Physically unjustifiable NOAA GHCN adjustments
The following is an excerpt from the thread
“In total, I have found 194 instances of WMO stations where “cooling” has been turned into “warming” by virtue of the adjustments made by NOAA to the raw data. As can be seen from the following “Cooling turned into warming” table (Table 1) below, which lists the Top 30 WMO station on the “cooling to warming” list, Darwin is ranked in only 26th place! The list is sorted by the absolute difference in the magnitude of the raw to adjusted slopes i.e. the list is ranked so that the worst case of “cooling” converted to significant “warming” comes first, followed by the next worse etc.
It’s clear from looking at the list that Darwin is certainly not “just a special case” and that in fact that there are many other cases of WMO stations where (as with Darwin) NOAA have performed physically unjustifiable adjustments to the raw data. As can been seen from Table 1 many of these adjustments result in trend slopes which are greater than the IPCC’s claimed 0.6 deg. C/century warming during the 20th century said by the IPCC to be caused by man’s emissions of CO2 through the burning of fossil fuels.
”
KevinUK