Jennifer Marohasy

a forum for the discussion of issues concerning the natural environment

A Note on Temperature Anomalies by Tom Quirk

By

One of the most vexing things about climate change is the endless debate about temperatures. Did they rise, did they fall or were they pushed? At times it seems like a Monty Python sketch following either the Dead Parrot or the 5 or 10 Minute Argument.

However it is possible to see some of the issues by looking at the correlation of the five temperature series that are advanced by the uppers or the downers.

The five groups are:
1. GISS, The Goddard Institute, home of James Hansen,
2. NCDC, The National Climate Data Center, a part of NOAA (as is GISS), the National Oceanographic and Atmosphere Administration.
3. BMO/UEA, The British Meteorological Office and the University of East Anglia.
4. UAH, The University of Alabama, Huntsville, home of Roy Spencer with his colleagues including John Christy of NASA and
5. RSS, Remote Sensing Systems in Santa Rosa, California, a company supported by NASA for the analysis of satellite data.

The first three groups use ground based data where possible with a degree of commonality. However since 70% of the surface of the earth is ocean and it is not monitored in a detailed manner, various recipes are followed to fill the ocean gap, if that is the best way of putting it.

The last two groups use satellite data to probe the atmosphere and with the exception of the Polar Regions which are less than 10% of the globe, they get comprehensive coverage.

One question is of course are the two groups measuring the same temperature? After all the satellite looks down through the atmosphere, while the ground stations are exactly that.

One of the ways to probe this is to look over time at the degree of correlation achieved in the measurements of the “global temperature anomaly

The results of such a comparison are given in Table 1 for the monthly time series from 1979 to 2008. There is the Pearson correlation coefficient extracted from the data. A value of 1.00 shows the compared values move in step with each other while a value of 0.00 would give complete independence. (A value of-1.00 is also possible.) “Commonality”, the square of the correlation coefficient is interpreted as showing what proportion of one measurement series is covered by the other series. Note that correlation does not imply connection or causality except that we know there is some commonality with ground based measurements.

Table 1.
Tom Quirk_table1_temp.JPG

First a check of the land based measurements shows that two groups are closely aligned, the difference reflecting the different processing to get the global result.

GISS is more problematic with less commonality which must be a reflection of quite different processing assumptions to that of NCDC or BMO/UEA.

For land based measurements we are faced with a “Judgement of Paris” and it is not clear who gets the Golden Apple.

Finally the satellite measurements have a high internal commonality but a commonality of some 50% with the land based measurements.

None of this should be surprising. The land measurements are on the land and subject to a number of uncertainties, such as heat island effects and lack of extensive ocean measurements while the satellites probe the atmosphere but not ground level.

So for the last 8 years the results are in Table 2

Table 2.
Tom Quirk_table2_temp.JPG

It is surprising to see the agreement achieved by two quite independent approaches.

However we should be aware that none of this is simple.

Tom Quirk
Melbourne

Reader Interactions

Comments

  1. spangled drongo says

    Has any long term testing on, say, a coastal area away from UHI effect, ever been carried out by both a satellite and human measured ground station to compare differences?
    The more remote systems become involved, the more error appears.
    EG Matthew Flinders’ tide predictions were spot on – BoM’s can be 2 hours out.
    Can’t help being sceptical.
    I couldn’t see anything on Anthony Watts blog.

  2. cohenite says

    That’s just great; not to mention the ‘debate’ about an average global temp and the fracas about base periods; speaking of which, can Tom respond to the unbelievable flack his and John McLean’s paper on the Climate Shift receives; as someone observed, with the temp step-up contemperanous with this ‘event’, and the increases in upward trends that base periods, either 61-90, or worse, 51-80, obtain as a result of the ‘event’, you’d think this would get more positive publicity; but apart from Tamino calling it a ‘crackpot’ theory and Joe D’Aleo doing a good piece on it, it doesn’t seem to receive much attention; but if it is right, it alone puts the kibosh on AGW temp increases. The D’Aleo piece is here;

    http://climatepatrol.org/2008/04/16/hatcrut3-global-land-ocean-temperature-index-1850-2008/

  5. cohenite says

    luke; we may be at cross purposes here; I’m aware of the PDO/IPO climate fluctuations so capably enunciated by Stewart Franks; I’m more interested in that described partial cessation of the pacific upwelling;

    http://mclean.ch/climate/Aust_temps_alt_view.pdf

    MY understanding was this was a oncer, not related to PDO periodicy, even though McLean and Quirk use SOI to measure its influence and effect; are you saying the alteration in upwelling is a feature of PDO switch?

  6. Jennifer says

    What about focusing on the information at hand? Usually there is debate about how the GISS data differs from the UAH data. But I think Tom makes a good point that in the scheme of things they are not that far appart?

  7. Luke says

    On (1) – what you’d expect when a warming signal involving the Arctic appears in recent years. i.e. GISS includes the Arctic – CRU does not.

    on (2) – I don’t really know but I find McLean very hard to take seriously. From memory his case is based on 3 references which he quoted here in some famous stoush.

  9. cohenite says

    Well the most weird thing is that land-based BM shows a change of-.9, the second highest in -ve trend terms; the av over the lot is -.32, and only GISS shows a +ve trend; all of them have error bars far greater than the revealed trends; so I don’t know about a judgement of paris; more like a judgement of clem kadiddlehopper. Really, how could you base major economic reform on this lot

  10. gavin says

    Jennifer: Don’t get me wrong as it’s good to have a robust debate over methods of analysis but in the end we are only fiddling with figures, not actual circumstances.

    For example; while looking at cohenite posts on the Booker thread, re historic observations at Bourke, Broken Hill and Wilcannia I couldn’t help thinking the emphasis on temp differences even rainfall ignore things like R/H, due point and in particular wind speed over soils etc.

    IMO drought is best read from historic trees, dam levels and other natural yields.

  11. cohenite says

    gavin; a large number of those data histories have wind speed information; generally you are correct; climate and its impact is more than a couple of prominent indices; the fact is though, AGW has hitched its star to temp, and now that temp is problematic, the complexities and stochastic elements of climate are being wheeled out; Hurst has surfaced here, and I note Tamino is having a stab at it.

  13. gavin says

    Cohenite: “AGW has hitched its star to temp (1), and now that temp is problematic, the complexities and stochastic elements of climate are being wheeled out (2)”

    (1) Not Me! The problem I have is many of the internet commentators IMO have never done a field measurement but they think they can hang their hat on the accuracy (or otherwise) in someone else’s work

    (2) When measuring in turbulence it was always thus. Only bloggers have yet to come to grips with that hey

    BTW Luke; in years of temperature measurement I never used a Stevenson screen and I was often up to my neck in UHI. Imo these issues are only man made for the uninitiated.

  14. John says

    Yes “cohenite”, Tamino did try to claim that the paper that I wrote with Tom Quirk was a “crackpot” theory. If my recollection is correct – and I admit that I don’t pay a lot of attention to professional “defenders of the faith” – he failed to present any decent argument against it and in fact seemed rather confused. All rather odd given that the Great Pacific Climate Shift of 1976 is well acknowledged by professional climatologists.

    As regards Tom Q’s latest as presented here, I have to say that the Hadley folk (i.e Phil Jones et al) appear to be more careful about their data of late.

    That said, I am aware of instances where the relationship between the data from 3 of adjacent grid-cells of their dataset very abruptly shifted, and other cases where it is obvious that the data was misread (eg. a 4 may have been take for a 9, a negative sign missed). My paper on the subject is in preparation.

    The other question of course is whether urban heat island effects might not also exist in temperature data of the lower troposphere. It seems very plausible that heat output over densely populated regions would impact the troposphere. Perhaps Europe with its stable population is unlikely to generating a lot more heat than 15 years ago but China certainly might be. And the question must be asked about how this heat is dispersed.

  15. Luke says

    So strange though that the Great Shift would be missing from the mainstream literature. (unless you want to count rags like E&E)

  16. spangled drongo says

    It was during this ’76-’81 climate shift that much of the Coral Sea cyclone activity moved to the Arafura Sea which has severely effected sub-tropical Qld’s weather ever since and no doubt extends further than that.
    I’ve said to Luke before that this was a significant and sudden climate happening and has no correlation with progressive ACO2 emissions.

  17. Alarmist Creep says

    Could just be relatively more Los Ninos than Las Ninas – tropical cyclones track further out in the Coral Sea and central Pacific in El Nino years and tend not cross the Queensland coast. Who needs a mysterious unpublished “shift”.

  19. lucia says

    Tom–
    I’ve been looking at these also, and organizing it differently. I’ve assumed the commonality is due to real changes in the earth’s temperature. To estimate that, I take the simple average of all five measurement systems.

    I then did a correlation analysis on the differences between instruments and that average.

    Using data from Jan1, 2001-March 2008, I get the following correlation matrix.
    —-
    GISS HadCRUT NOAA UAH MSU RSS MSU
    GISS 1.000 -0.123 0.274 -0.290 -0.688
    HadCrut -0.123 1.000 0.500 -0.816 -0.297
    NOAA 0.274 0.500 1.000 -0.739 -0.764
    UAH -0.290 -0.816 -0.739 1.000 0.504
    RSS -0.688 -0.297 -0.764 0.504 1.000
    GISS HadCRUT NOAA UAH MSU RSS MSU

    —-
    (Of course, it’s symmetric. Also… I’ll admit I haven’t doubled checked before posting in a comment.)

    Because the land based systems do share data, I expecting the land instruments to be positively correlated with each other, and negatively correlated with the satellites. That is almost the case, but GISS and HadCrut show negative correlations.

    I don’t know the significance of this. It’s possible there is no significance, except it wasn’t quite what I expected.

    Given what this operation does, I was expecting to see the satellites show positive correlations with each other and negative with the land instruments. That happens.

    I may look at longer time frames, as you did. (My main interest was estimating the uncertainty due to measurement. The standard deviation about the mean at any time is roughly 0.05C. This suggests a lower bound for monthly “measurement noise” of about 0.05C. )

  20. Tom says

    Lucia

    My initial comment is that using an average might get you into trouble as you are already using the data you intend to test. Strongly correlated and connected series (say using the same data) will not improve the analysis but rather distort the result.

    It is cleaner to compare individual time series.

    I guess that the commonality of the results, correlation squared, is a reflection of real temperature fluctuations but it would also carry common distortions shared by say the land based system.

    This may not help but let me think some more about it.

    Tom

  22. lucia says

    Tom–
    Yes. I agree the commonality between the two data sets is due to both:
    a) the fact they intend to measure the same thing (the GMST in a particular year) and
    b) the fact that the different groups both sometimes rely on the exact same measurement.

    With regard to (a), the commonality should be a sign the groups are succeeding at measuring the GMST. Ideally, if all groups measured GMST perfectly accurately and precisely, but by independent methods, the correlations between measurements would be 100%.

    With regard to (b), it happens that sometimes the groups use the exact same measurement, from the same thermometer at the same station. In this sense, some of the commonality in official records is simply due to using the same measurements.

    It’s a bit as if I downloaded GISS twice and averaged it to get “new” data, and then thought I’d actually used 2 data sets. That’s not right because the 2 data sets are really just two copies of 1 thing, and the copying includes the exact same errors.

    To the extent that the different agencies methods use different data or methods of massaging data, comparing the differences gives us an estimate of measurement errors associated with determining the GMST. However, this always gives us a lower bound because, to the extent that the agencies share some raw data when calculating their averages, the difference between the two data sets is smaller than would exist if they actually measured things independentlly.

  23. Sam says

    Wow, I’ve read a lot of nonsense from Gavin and Luke on this blog but this by Gavin has to take the cake: BTW Luke; in years of temperature measurement I never used a Stevenson screen and I was often up to my neck in UHI. Imo these issues are only man made for the uninitiated.

    This is absolutely breathtaking. UHI only has a man made component for the uninitiated (as opposed to being initiated into what – the grass skirt society?)

    So something (UHI) which by definition is caused by man, only has a man made component for those of us among the unwashed who aren’t as smart as Gavin. But then, everything else involved with “changing the climate” is caused by man.

    Do you people even think about what you are saying or do your lips just start moving?

  24. Tom says

    Lucia

    Absolutely correct in your analysis.

    You can also see the “double counting” if you average NCDC and BMO together and then compare with GISS. NCDC and BMO combined standard deviation is no better than the separate comparison.

    Tom