Independent scientists to referee global temperature adjustments

terence kealey

Professor Terence Kealey.

Source: Inquiry Launched Into Global Temperature Data Integrity

London: 26 April 2015. The London-based think-tank the Global Warming Policy Foundation is today launching a major inquiry into the integrity of the official global surface temperature records.

An international team of eminent climatologists, physicists and statisticians has been assembled under the chairmanship of Professor Terence Kealey, the former vice-chancellor of the University of Buckingham.

Further details of the inquiry, its remit and the team involved can be seen on its website at

The NZ Climate Science Coalition (NZCSC) intends to submit to the inquiry material about the New Zealand temperature record produced by NIWA, its adjustments and a review of the NZTR by Coalition scientists.

I hope we emphasise that NIWA has always refused any genuinely independent review of the NZTR. The only ‘peer review’ NIWA has published was no more than an in-house review by their mates at the Australian BOM which disturbed not a single item of data or methodology. In other words, it was furtive, shallow and firmly non-invasive—think x-ray examination by a blind radiologist.

We cannot ignore NIWA’s extraordinary assertion to the High Court review that NIWA has “no obligation to pursue excellence,” nor their insistence that there is no such thing as the official New Zealand temperature record.

Aside from that, we wish this blessed inquiry well.

18 Thoughts on “Independent scientists to referee global temperature adjustments

  1. Alexander+K on April 28, 2015 at 8:48 am said:

    This new initiative looks positive!

  2. Richard C (NZ) on April 28, 2015 at 10:33 am said:

    Christopher Booker’s article on this inquiry in the Telegraph features the GISS adjustments to Puerto Casado, Uruguay:

    ‘Top scientists start to examine fiddled global warming figures’

    Puerto Casado is one of 3 stations GISS smeared all over central South America to get the anomaly that contributed most to their 2014 “warmest ever” year. Except when you look at the the GISS adjustments to Puerto Casado vs the BEST adjustments to the same station it becomes abundantly clear that the adjustments are shonky, to say the least.

    Paul Homewood at notalotofpeopleknowthat was the first to alert everyone to GISS Uruguay Paraguay, looks like Booker was following the trail. In the Temperature Records Open Thread in this blog (CCG) this situation is recorded and there is a cross link to the post ‘Scandel heating up’ where there is more detail:

    Covers GISS and BEST adjustments at Puerto Casado, Paraguay and Gisborne Aero, New Zealand along with links to the Climate Etc discussion of same (extensive, 1000+ comment thread) and a bunch of other stuff. The results from the respective adjustment methodologies are not consistent and produce some massive step changes but not from the local station data and history i.e. the adjustments are imputed from remote locations or even by GISS GCM (CO2 forced – see Gisborne Aero below) but there is no local justification.

    If the inquiry can’t get to grips with the situation just by Puerto Casado alone then they haven’t a hope of sorting this out globally.

    The following, Gisborne Aero NZ, is an example of GISS CO2-forced model-based adjustments to historical temperature records:

    Gisborne Aero monthly plotted on this page (click graph to zoom in):

    There is no reason for a break between 1974 and 1979. But 1975/76 is only a 0.1 adjustment. There is a progressive cumulative adjustment in 0.1 increments adding to 0.7.

    GISS raw monthly data (as plotted):

    GISS adj monthly data:

    See metANN column at far right of the data sheets.

    At 1963 the cumulative adjustment is 0.7
    At 1968 the cumulative adjustment is 0.6
    At 1972 the cumulative adjustment is 0.5
    At 1975 the cumulative adjustment is 0.4
    At 1980 the cumulative adjustment is 0.3
    At 1982 the cumulative adjustment is 0.2
    At 1986 the cumulative adjustment is 0.1
    At 2001 the cumulative adjustment is 0.1
    At 2002 the cumulative adjustment is 0.0

    There is no valid reason for adjustments of this nature. And there is no resemblance to the BEST adjustments:

    Duplicated at Paul Homewood’s.


  3. Richard C (NZ) on April 28, 2015 at 10:44 am said:

    >”looks like Booker was following the trail”

    Actually, now that I remember better, Booker has been all over it. Booker:

    “Back in January and February, two items in this column attracted more than 42,000 comments to the Telegraph website from all over the world. The provocative headings given to them were “Climategate the sequel: how we are still being tricked by flawed data on global warming” and “The fiddling with temperature data is the biggest scientific scandal”. ”



  4. Richard C (NZ) on April 28, 2015 at 11:01 am said:

    Frankly, got my doubts that the inquiry team has the in-depth knowledge of the respective homogenization and adjustment methodologies of each group (GISS, BEST, BOM, etc) to identify shonky adjustments, let alone the the arbitrary adjustment application issues of particular groups e.g. GISS.

    I just hope they don’t blow it. We’ve already had a local and “unnecessarily prolix” court case failure. A global inquiry that doesn’t get to the critical issues would be an immense and embarrassing failure on behalf of all those who have been digging into the nitty gritty and have the understanding. Not to mention the setback such a failure would be – imagine the mileage all the warmies would make of it.

  5. Richard Treadgold on April 28, 2015 at 11:05 am said:

    “Frankly, got my doubts that the inquiry team has the in-depth knowledge”

    Exactly, which is why they’ve already requested contributions from those who do. Sounds like there could be a bit of collating and writing in your future… 😉

  6. Richard C (NZ) on April 28, 2015 at 11:14 am said:

    From link:

    Terms of reference

    The panel is asked to examine the preparation of data for the main surface temperature records: HadCRUT, GISS, NOAA and BEST. For this reason the satellite records are beyond the scope of this inquiry.

    The following questions will be addressed.

    1. Are there aspects of surface temperature measurement procedures that potentially impair data quality or introduce bias and need to be critically re-examined?

    2. How widespread is the practice of adjusting original temperature records? What fraction of modern temperature data, as presented by HadCRUT/GISS/NOAA/BEST, are actual original measurements, and what fraction are subject to adjustments?

    3. Are warming and cooling adjustments equally prevalent?

    4. Are there any regions of the world where modifications appear to account for most or all of the apparent warming of recent decades?

    5. Are the adjustment procedures clearly documented, objective, reproducible and scientifically defensible? How much statistical uncertainty is introduced with each step in homogeneity adjustments and smoothing?

    # # #

    Of these I think 5. is (most) critical. Also problematic in terms of procedure and results comparisons e.g. GISS vs BEST.

  7. Richard C (NZ) on April 28, 2015 at 11:19 am said:

    >”Sounds like there could be a bit of collating and writing in your future…”

    Not until after kiwifruit post harvest unfortunately. I’m working 10.5 hr nightshifts which doesn’t leave much time for anything else, doesn’t leave a clear head either.

  8. Richard Treadgold on April 28, 2015 at 11:31 am said:

    That’s a tough regime, all right. But your contributions here don’t appear to suffer; they are outstanding.

  9. Richard C (NZ) on April 28, 2015 at 12:08 pm said:

    Just recalling what’s already been raked over RT, nothing new. Writing it up in a coherent report scares me witless, there’s so many rabbit holes to go down but what’s really worth communicating?

    Example follows of the difficulty of communicating the issues and another of BEST’s avoidance of same.

  10. Richard C (NZ) on April 28, 2015 at 12:16 pm said:

    The following was studiously avoided by Zeke Hausfather (BEST, along with Mosher) at Climate Etc:

    Zeke, this would apply directly to BEST would it not?

    ‘Circularity of homogenization methods’

    by David R.B. Stockwell PhD, October 15, 2012

    The proposition is that commonly used homogenization techniques are circular — a logical fallacy in which “the reasoner begins with what he or she is trying to end up with.” Results derived from a circularity are essentially just restatements of the assumptions. Because the assumption is not tested, the conclusion (in this case the global temperature record) is not supported.

    I present a number of arguments to support this view.

    First, a little proof. If S is the target temperature series, and R is the regional climatology, then most algorithms that detect abrupt shifts in the mean level of temperature readings, also known as inhomogeneities, come down to testing for changes in the difference between R and S, i.e. D=S-R. The homogenization of S, or H(S), is the adjustment of S by the magnitude of the change in the difference series D.

    When this homogenization process is written out as an equation, it is clear that homogenization of S is simply the replacement of S with the regional climatology R.

    H(S) = S-D = S-(S-R) = R

    While homogenization algorithms do not apply D to S exactly, they do apply the shifts in baseline to S, and so coerce the trend in S to the trend in the regional climatology.

    The coercion to the regional trend is strongest in series that differ most from the regional trend, and happens irrespective of any contrary evidence. That is why “the reasoner ends up with what they began with”.

    I’ve sought clarification from you upthread on the BEST process upthread here:

    Nothing forthcoming addressed to me but you did state this:

    Zeke Hausfather | February 11, 2015 at 12:26 pm |
    A slight correction: I should have said “These records are combined by aligning the mean values of each subsegment relative to the regional expectation (e.g. based on comparisons to nearby stations) of the station record, as shown in the example above”

    Mosher states:

    “if a station that was say 1C offset from all its neighbors
    suddenly becomes and stays .75C offset, then it is split.”

    My initial request for clarification was accompanied by reference to a conventional break adjustment method (R&S93) and a controversy in respect to that (NZCSC v NIWA) that considers the data for the 24 months on either side of the break and statistical criteria for an accept/reject decision on an adjustment.

    The BEST process appears to have dispensed with all that, instead, as hidethedecline puts it upthread, “regional expectations are the primary parameter” of the BEST process.

    I dispute the BEST process as described above and elsewhere in this thread if it makes no recourse to the local data either side of the break. And in particular the use of the mean of a segment rather than the data adjacent to a break. This I support by initial case study of the 1971 break at Puerto Casado, Paraguay:

    Segments 1951-1971 (1) and 1971-2006 (2):

    The adjustment to the segment (1) mean appears to be 0.8 and GISS makes a 0.78 adjustment at the same break. Segment (2) is on the anomaly baseline relative to “regional expectation” (near enough). I repeat from my initial comment linked above:

    “The mean difference between (1) and (2) is about 0.8. If this adjustment has actually been made it is non-trivial i.e. 0.8 is a massive adjustment requiring a rock solid basis. That basis is the station history, local factors, the statistical methodology and criteria, etc, all of which must be able to be dissected step-by-step.”

    Apparently BEST ignores all of the local considerations in favour of “regional expectation”. I call bogus. One of the local considerations is the data adjacent to the break either side which is a conventional and established methodology. 24 months either side is generally statistically adequate (R&S93, follow link to initial comment above). If you look at the 2 yrs either side of 1971, THERE IS NO MISMATCH i.e. the data certainly does not “suddenly becomes and stays [0.8]C offset” as Mosher states.

    The mean of segment (1) and the 0.8 (1) – (2) difference is established by data 5 to 15 years prior to the break where the average level is higher than at the break. But that data is irrelevant to the break. The average level of the segment (1) data has reduced to be the same as that of segment (2) on the other side of the break over the 5 years prior to the break.

    Therefore, the 0.8 BEST and GISS adjustments to Puerto Casado, Paraguay, 1951-1971 data are invalid. As is the BEST process in this case.

  11. Richard C (NZ) on April 28, 2015 at 12:28 pm said:

    Example of the difficulty of communicating the critical issues with respect to Puerto Casado via the ‘Scandel heating up’ link upthread to Climate Etc:

    Me replying to Brandon Schollenberger:


    >”. BEST can find a breakpoint by examining a 10 year period even if that 10 year period is within a 40+ year period it doesn’t find a breakpoint in. BEST didn’t just compare 1951 on when looking for that breakpoint. It compared 1951 on, 1952 on, 1953 on, etc.”

    Firstly, be careful not to conflate break detection with subsequent adjustment accept/reject criteria.

    Secondly. 1971 is NOT an “Empirical break”, it is a “Station Move”:

    A station move is usually determined from the station history. I don’t know how BEST identifies station moves such as this but I don’t think it is from the station history i.e. their primary detection is by automated break-point analysis. Why and how a break detected in this way can be attributed to a station move without recourse to the station history is a mystery to me. But I’d like to be enlightened.

    Thirdly, The length of the period prior to the breakpoint is VERY relevant to the adjustment process in this case. Look at the reconstructed series above, the mean of the segment prior to 1971 (the target series) is the mean of the ENTIRE 20 yrs of the segment – NOT the 5 years that would be used by a 10 yr break detection period. And I repeat, 2 yrs of monthly data either side of a break is sufficient for statistical adjustment accept/reject criteria.

    >”I believe the minimum period BEST can use when searching for breakpoints is 10 year. If so, they only need 10 years of data around a point to decide there’s a brekapoint there. That means you’d need to include almost any stations which have data for the year of the breakpoint”

    Fine. That’s to detect a break in the target series (and see below re the comparator data adjacent to the break). But that’s NOT how BEST make the adjustment. To do that they split off the ENTIRE segment (20 yrs in this case – NOT 5 yrs), take the mean, and that’s the target data for the “new” station. So the comparator data must correspond to the 20 yrs of target data 1951 – 1971. It’s the corresponding comparator data that I’ve identified upthread from the “nearby” stations. And it’s not good. Just for starters, Pora Pora has a break right in the middle of the corresponding segment.

    And I hope you’ve taken a long hard look at the comparator data where local breaks occur right next to the 1971 target break but on either side i.e. conventional cross-break statistical analysis might turn up some interesting results. One of the possibilities being that the 1971 break is not actually a break because it was common to the neighbours.

  12. Richard C (NZ) on April 28, 2015 at 12:45 pm said:

    >”follow link to initial comment above”

    The “initial comment” (linked above) addressed to Zeke Hausfather was this (which was also sidestepped by Zeke):

    richardcfromnz | February 11, 2015 at 7:17 pm |

    [A] >”Technically the series is cut at breakpoints and every segment is treated as an individual station for the purposes of constructing the underlying temperature field. However, the general point is that it is the mean temperature of the segment, rather than its start or endpoints, that is relevant when combining it with other stations to estimate the temperature field.”

    [B] >”We also produce “adjusted” records for each station, though these are not actually used for the Berkeley temperature product and are solely for those interested in data from that specific station. These records are combined by aligning the mean values of each subsegment of the station record, as shown in the example above.”

    OK, case study. 1971 break PUERTO CASADO, Paraguay:

    Segments 1951-1971 (1) and 1971-2006 (2):

    I assume that segment (1) in [A] is an “adjusted” series, but an inference can be drawn from [B] that it is NOT adjusted, simply separated from (2). Can you clarify EXACTLY what the process is please Zeke?

    The mean difference between (1) and (2) is about 0.8. If this adjustment has actually been made it is non-trivial i.e. 0.8 is a massive adjustment requiring a rock solid basis. That basis is the station history, local factors, the statistical methodology and criteria, etc, all of which must be able to be dissected step-by-step.

    BOM was caught out by this at Rutherglen. They relied on automation to detect an empirical break. When questioned they couldn’t say from the local station history why the break occurred i.e. they had neglected human input in their process. They were eventually able to deduce from the records that the site moved from the north side of the rise to the south side. I accept their reasoning. But I don’t accept their adjustment. Max stayed exactly the same but Min changed by over 1.5 I think is was. The adjustment was on the basis of other sites, not the local conditions. A change in Min of that size from one side of a rise to the other when Max remains the same is highly suspect. The methodology could be proved, or otherwise, simply by re-installing an AWS site on the north side and taking at least a years data. Nowhere has this ever been done to my knowledge.

    [B} definitely states an “adjustment” which I estimate to be about 0.8 above. GISS makes a 0.78 adjustment at the same break. I don’t question the process of adjustment for site moves if the station history details the move or it can be ascertained as at Rutherglen. I do question the rationale of “empirical break” adjustment without recourse to station history. And I certainly question your rationale for a 0.8 adjustment at 1971 Puerto Casado, especially if the adjustment is for [A] in addition to [B] and if you are in fact using segment means.

    Surely you are not using the entire respective segment means (1) and (2) to arrive at the 0.8 step? That is ludicrous if so. This is the central issue in the New Zealand 7SS controversy that went to court (NZCSET v NIWA). The Judge’s duty was not to decide questions of science but to decide questions of fact. He failed to do his duty. The NZCSC rigourously followed the established method of Rhoades & Salinger (1993) in compiling the 7SS series, NIWA departed arbitrarily from the R&S93 method so that they cannot now cite the basis of their method. NIWA’s 7SS is the NZ data for CRUTEM4/HadCRUT4. The NZCSC method has now entered the literature (De Freitas, Dedekind, and Brill, 2014) despite the Judge’s decision against NZCSET.

    Point is, the R&S93 method uses statistical accept/reject criteria for k = 1 and k = 2 i.e. 12 and 24 months either side of the break, 1 and 2 years of monthly data. Puerto Casado above is monthly data. Clearly 2 years either side of the 1971 break does not support a 0.8 adjustment. Or any adjustment, the data matches over the 4 year overlap. There was no adjustment in (2) for the similar break at 1987. The segment means are irrelevant.

    The statistical accept/reject criteria for R&S93 is in the Appendix here:

    ‘Statistical Audit of the NIWA 7-Station Review’

    Cross-posted at Climate Conversation Group here:

  13. Richard C (NZ) on April 28, 2015 at 3:23 pm said:

    Nice 12 month smoothing (running average) applied to GISS graph at Gareth Renowden’s Hot Topic:


    Looks like REALLY scary warming in 2015 (although not so much in model terms). Rob Painting swallowed it hook line and sinker:

    Rob Painting April 27, 2015 at 9:04 pm

    “Of course a reversal could happen well before then.”

    It likely already has – hence the uptick in global surface temperature at the end of the above graph. The changes in ocean circulation bear all the hallmarks of the positive (warm) phase of the Interdecadal Pacific Oscillation (IPO). We’ll find out for sure soon enough.

    Yes we will find out Rob but I think you’ll have to wait a while. Here’s the 12 months of GISS data for 2009/10 and 2014/15:

    2009.33 0.59
    2009.42 0.62
    2009.5 0.66
    2009.58 0.61
    2009.67 0.64
    2009.75 0.58 <<<<< Min =
    2009.83 0.71
    2009.92 0.58 <<<<< Min =
    2010 0.65
    2010.08 0.74
    2010.17 0.87 <<<<< Max 1
    2010.25 0.82 <<<<< Max 2
    2014.25 0.72
    2014.33 0.78 <<<<< Max 2= (2014.33 contributes to the 12 month running mean plotted 2015.17)
    2014.42 0.61
    2014.5 0.5 <<<<< Min
    2014.58 0.73
    2014.67 0.81
    2014.75 0.77
    2014.83 0.63
    2014.92 0.73
    2015 0.75
    2015.08 0.78 <<<<< Max 2 =
    2015.17 0.84 <<<<< Max 1


    What "uptick"?

    Now adding RSS and 12 mean samples (centered 12 month smoothing) the GISS graph isn't dramatic anymore (the 2010 El Nino barely a hump and no 2015 "uptick"):

    Any 2015 "uptick" is entirely absent from RSS too and the 2010 El Nino is still a pronounced spike even after 12 month smoothing. In other words, GISTEMP is insensitive to ENSO activity. The radiosondes are as sensitive as the satellites and far more sensitive than GISTEMP, see HadAT2:

    Anyone, like Rob Painting, pronouncing an impending IPO warm phase from that smoothed GISS graph are only fooling themselves. As Rob is.

    And anyone picking GISS without recourse to any other datasets is myopic. As Gareth is.

    Gareth of course is of a want to fool as many others as he can i.e. he’s intentionally myopic. And quite successful in the case of the willing-to-be-fooled Rob Painting.

  14. Richard C (NZ) on April 28, 2015 at 3:39 pm said:

    Dang, graph link should be:

    >”Now adding RSS and 12 mean samples (centered 12 month smoothing) the GISS graph isn’t dramatic anymore (the 2010 El Nino barely a hump and no 2015 “uptick”):”

  15. Richard C (NZ) on April 28, 2015 at 4:02 pm said:

    [Rob Painting] >”The changes in ocean circulation bear all the hallmarks of the positive (warm) phase of the Interdecadal Pacific Oscillation (IPO)”

    ALL the hallmarks?

    HadSST3 NH vs HadSST3 SH:

    Appears to be a hallmark missing from the Southern Hemisphere (60% of total sea surface area). And the NH hallmark isn’t what it was either.

  16. Richard C (NZ) on April 28, 2015 at 4:31 pm said:

    >”2015.17 0.84 <<<<< Max 1"

    If you switch the GISS "Smoothing Radius" from 1200 km down to 250 km the anomaly drops from 0.84 down to 0.82.

    There is no option for 0 km Smoothing Radius. One wonders what the anomaly would drop to if there was.

  17. Richard C (NZ) on May 3, 2015 at 11:16 am said:

    ‘Unadjusted data of [USA] long period stations in GISS show a virtually flat century scale trend’
    [1900 – 2000]

Leave a Reply

Your email address will not be published. Required fields are marked *

Post Navigation