Skip to content


October 10, 2012

By Paul Homewood



The Hadley Centre introduced their latest global temperature dataset, HADCRUT4, last January, but only provided revised historical temperature data up to 2010. In the meantime they have only been issuing ongoing data under the old HADCRUT3 set.

However they have now brought out monthly figures for HADCRUT4 which take us bang up to date for August 2012. Let’s take a look then at the differences between the old and new versions.




Figure 1

Figure 1 shows the 12-month running averages for the two sets, which helps to take out some of the noise produced by monthly variations. From 1980 to 1998, the two follow each other pretty closely. From around 2005, however, they start to diverge quite noticeably.

A look at the numbers bears this out.


12 month average anomaly HADCRUT3 HADCRUT4 Diff
Dec 1980 0.08 0.09 0.01
Dec 1985 -0.04 -0.03 0.01
Dec 1990 0.25 0.29 0.04
Dec 1995 0.17 0.20 0.03
Dec 2000 0.27 0.29 0.02
Dec 2005 0.48 0.54 0.06
Dec 2010 0.48 0.54 0.06
Aug 2012 0.36 0.41 0.05


The difference has increased by 0.04C since 1980. This may seem relatively insignificant. but it scales up to 0.13C per century, which interestingly is exactly the same amount that the latest version of GHCN adjustments has added to the warming trend. Perhaps more significantly it altered the comparison with 1998, thus:-


12 month average anomaly HADCRUT3 HADCRUT4
Dec 1998 0.55 0.52
Dec 2011 0.34 0.40
Increase/Decrease -0.21 -0.12


The new version increases warming (or rather decreases cooling) since 1998 by 0.09C, a significant amount for a 13 year time span. Whilst the changes should not affect the trend in future years, they will affect the debate as to whether temperatures have increased in the last decade or so.

So what is different in the new version? The paper by Morice et al (2012) explains that the major change is the addition of extra stations in and around the Arctic. As the Arctic has warmed in recent years, this filling in of gaps has tended to push up the global average temperature anomaly. However, the paper also makes clear that “Measurement coverage in the Southern Ocean and the Antarctic remains sparse”

We already know from UAH satellite data that the Antarctic has been cooling since 1980. It would therefore seem that HADCRUT4 has introduced a bias of its own, by increasing coverage of the warmer Arctic but not doing the same in the cooler Antarctic. The maps below, from the Morice paper, show just how significant this introduced bias is.




The maps on the right show the station coverage for HADCRUT3 and HADCRUT4, for January 2005. The extra coverage in the Arctic for the latter is clear. However the large gaps in the Antarctic and Southern Oceans in HADCRUT3, where the anomalies are largely negative, still appear in HADCRUT4.

If you were looking for a way to create an artificial warming trend, I cannot think of a better way than this.



There is an even better map below, from one of Phil Jones’ presentations that shows where the new coverage is. There is none at all anywhere near the Antarctic.



  1. October 10, 2012 9:28 pm

    The computation of global temperatures will always be problematic. Much of the global surface is ocean and there are large land areas, such as Antarctica, Australia, Canada and northern Africa, that have sparse data coverage. Plus, stations have different periods of records with many regions not having any records of more than a century. Plus, there are the well-known actions of many meteorological authorites of “adjusting” the raw temperature data to remove perceived inconsistencies (from Alice Springs to Reykjavik)!

    Personally, I prefer to examine the long-term temperature trends at individual stations that have not been influenced by urban or industrial effects and which have operated at the one location over a century or more. If there is a global change in temperature trend this would be expected to be shown in most of the individual station analyses.

    My studies indicate that the temperature trends at a wide variety of locations worldwide have not changed over the period since the mid-1800s. Certainly there have been short-term trend changes, but these are seen to be cyclical about an overall increasing trend of less than 1degC – not surprising at all as we all know that the world is warmer now than in the Little Ice Age of around 300 years ago. The important conclusion, from my viewpoint, is that there is no evidence of abnormal temperature trends in recent decades.

    Examples of temperature trends at long-term rural stations in Australia cane be seen at:

  2. Doug Proctor permalink
    October 10, 2012 11:31 pm

    The consistent and on-going warming of recent records and cooling of older records is a remarkable thing, most remarkable because those who are responsible for the changes don’t seem bothered at all by the clear trends in what they say is supposed to be a neutral +/- operation.

    This reminds me of the Bank of Nova Scotia commercial that says “You’re richer than you think you are”, while all of us are going backwards with our stocks and our home values.

    Cognitive dissonance is a powerful force to drive you into finding reasons to continue to believe. And the smarter, more knowledgeable you are, the better you are at finding them.

  3. Brian H permalink
    October 11, 2012 6:44 am

    Artificial warming trends are SO much easier to find than real ones.

  4. November 3, 2012 9:53 pm

    @Brian Gunter I did much the same study here:

  5. December 7, 2012 9:56 am

    A few things.

    1. The CRUTEM4 vs CRUTEM3 comparison isn’t valid, because CRUTEM is the land-only part of HadCRUT – so of course there is no extra coverage in the southern Ocean!

    2. I can see several areas in the Southern Ocean where there is extra coverage in HadCRUT4 compared to HadCRUT3, both in 1886 and in 2005. Look, for example, south of Australia in 1886, or in the South Pacific in 2005.

    3. Using clever mathematics, I understand that you can estimate the uncertainty in the global temperature that is created by not having observations for the entire globe. These are available for the HadCRUT4 data. If you plot these on your graph of global temperature then it is clear that the observations simply sren’t good enough to say whether the globe has warmed or cooled over the last 10-15 years, but you can see that it has warmed over the last 50 years.

    One of the most fundamental lessons I was taught in A-level physics was to always show the error bars. This is what the Hadley Centre do. Why don’t you?

  6. Bob permalink
    January 28, 2014 1:12 pm

    Ah, I see. The temperatures were going down so the data set was changed to make it look like the world was warming.

    • tomwys1 permalink
      July 17, 2016 5:59 pm

      Keen observation, Bob! To see how they did it, google:

      • tomwys1 permalink
        July 17, 2016 6:00 pm

        The Problematic Transition to HadCRUT 4 from HadCRUT 3


  1. HADCRUT4 hates 1998 « sunshine hours

Comments are closed.

%d bloggers like this: