GHCN V3 Update and Comparison

On Nov 22nd I found a problem with 3 of the months of the new GHCN V3 anomaly data available from the NCDC.  It appears that the problem was caused by data being missing for certain years for particular months.  This led to the data being offset and this caused the set difference stated for the data to be incorrect.  As pointed out by other sites this is the type of quality control that is painful to miss.

Now that the problem has been corrected it is possible to take a look at the difference between V2 and the new V3.  The results are also interesting, but for a different reason, especially if someone were Skeptical about the goals of the new version.  Here is the annual difference between the sets from 1880-present.

The Inconvenient Skeptic

GHCN Version Differences on an Annual Basis

Not surprisingly the new version has higher temperatures on an annual basis for every year since 1953.  Since 1978 the average annual difference is 0.18 °C.  That means that if the new version was put into place as the default, it would shift the warming since then by that amount each year.  Even more frustratingly is the period from 1904-1940.  That period has an average difference of -0.16 °C.  So just the change in versions generates the appearance of 0.34 °C more warming than V2 shows.

If the accumulated difference is shown, the difference is even more pronounced.  From 1953-2010 there is more than 8.0 °C accumulated difference, just between the versions.  Changing versions will result in the appearance of significantly more warming than V2 has shown.  Considering that V2 is already one of the warmest versions, the change will be significant.

The Inconvenient Skeptic

GHCN Accumulated Difference

Sadly this is not a surprise.  That an updated version would show enhanced warming is predictable at this point in the debate.  Version 3 will in effect shift the entire temperature record for the past 67 years upwards.  That it also decreases the temperature of the period before that only enhances the effect.  This is a subtle change, but one that strongly favors the appearance of global warming.

Posted in Anomaly by inconvenientskeptic on November 25th, 2010 at 4:45 am.


This post has 8 comments

  1. Malaga View Nov 25th 2010

    Sadly this is not a surprise.
    It really is a very sad state of affairs… these guys are without honour, integrity or shame.

  2. Rational Debate Nov 25th 2010

    Just love how they warm the little ice age, cool the warm 1910-1940’s, warm since then…. more hockey schtick treatment of temperatures.

    Its going to be interesting to see how they justify these changes compared to V2.

    Thanks for your work first finding the problems, then catching the differences here!

  3. cthulhu Nov 25th 2010

    What on earth is this?

    The annual difference you’ve *added* the monthly differences for each year and *forgotten* to divide by 12 (there are *12* months in a year) to get the annual difference.

    Divide every number by 12 in the first graph.

    The second graph is just off the wall bizzare. Why on Earth would you accumulate the differences? The differences don’t accumulate.

    Do you really think V3 shows 8C more warming in the last 60 years than V2?

    Sadly this is not a surprise. That skeptics are too incompetent to analyze data correctly is predictable at this point in the debate

  4. inconvenientskeptic Nov 25th 2010

    You would be correct if I was trying to do a statistical analysis between the two sets. Statistically they are nearly identical. What is interesting is the trends in the difference. The accumulation chart is effective at showing when the trend in the difference. When V3 is primarily less than V2, the slope is negative. When V3 is greater than V2, the slope is positive.

    That the slope is positive for that past 57 years indicates that V3 shows more warming than V2 from that point forward. That the 40 years from 1900-1940 showed steadily lower temps for V3 goes to show that V3 will give a subtle effect of making the period after 1953 appear warmer than the period before 1940. If each month was plotted it would almost be impossible to see this difference because the average difference for each month is only 0.00286 C.

    When dealing with such subtle difference unconventional means are needed. Saying that the average difference from 1953-2010 is 0.02367 C greater than the average difference from 1900-1940 is not an effective method of describing the effect of the new version. I could also point out that the difference between 1998-2010 and 1900-1940 is 0.02784 C. That still doesn’t express that every year since 1953 shows more warming for V3 than V2. That the difference increases as it approaches the current day is certainly curious.

    If the new version generated a random difference, the accumulation would be zero. More than anything the accumulation indicates that the difference between the sets is not random.

  5. Malaga View Nov 26th 2010

    the accumulation indicates that the difference between the sets is not random.

    That is exactly what I understood from the accumulation chart… and sadly for cthulhu the skeptics are too incompetent at analysing the data.

  6. cthulhu Nov 26th 2010

    “I could also point out that the difference between 1998-2010 and 1900-1940 is 0.02784 C. That still doesn’t express that every year since 1953 shows more warming for V3 than V2. That the difference increases as it approaches the current day is certainly curious.”

    The first graph tells you that. The accumulation graph is just nonsense. Furthermore if the Y axis was correct on the first graph it would also show the difference between 1998-2010 and 1900-1940 was about 0.03C.

    And everyone could go “So?” Because 0.03C out of the total warming of about 0.8C is irrelevant. Smaller than the uncertainty.

    That it’s not random is to be expected. The V3 adjustment method is different than the V2 adjustment method. As the adjustment methods involve changes to trends it’s highly unlikely the V2 to V3 differences will all be distributed randomly and the trend will remain the same. But 0.03C difference is hardly enough to warrant getting knickers in a twist.

    I like how Watt’s exploited your earlier post to smear GHCN V3. As if an error on the website meant there was a problem with the quality of the underlying data…

  7. inconvenientskeptic Nov 26th 2010


    Interestingly enough, the 0.8C is statistically irrelevant as well. What is frustrating is that every revision or update slowly, but surely shows more warming than the one before. The difference between the satellite and the station data keeps increasing in this manner. That the different methods are diverging with each update is a real problem. That the new version “only” shows 0.03C more warming than the one before is the point. It subtly lowers the temperature 100 years ago and increases the modern temperature.

    As for Watts. His point was quality control. There are problems with quality control in the station data. This was simply another problem in quality control. His point (and mine) is that we don’t know if there are other errors in the underlying data. Since every revision takes the data farther away from the more comprehensive satellite data I would argue that there is a problem with the underlying data. The satellite data is accurate enough to produce 10 day forecasts for most of the Earth that are reasonably accurate. The station data doesn’t provide the coverage needed for accurate forecasts. The only reason to look at the station data today is because that is all there was prior to 1979. All the effort with station data should be to correlate it to the satellite data. Not to take the station data seriously.

    I am more irritated that there isn’t already a standard method for using all the available methods of measuring temperature into an open, standard temperature. I want one standard one so I can spend my time analyzing data instead of sorting it. The checks I run when I finish sorting are what caught the errors. The accumulation is one I use to catch small systematic errors. Clearly that quality control method works because it caught out that the website data was not properly aligned. So even if you don’t think it is useful, it did what it was supposed to do. Similar things are done with the ice core data as well.

    I do appreciate your comments, even if we disagree. I prefer having people with both views express their ideas. Thanks for doing that.


  8. You’re to be congratulated for exposing the significant changes introduced into the entire data base by the GHCN data keepers. The most odious aspect–the suppression of the previous peak near mid-century–is illustrated nicely by the cumulative graph. This smacks of the same motives as attempts to downplay the MWP in the proxy data: hide the natural cycles while manufacturing a trend. The idea of data being the product of measurements has been debased by “climate science.”

Web Design & Dev by

Mazal Simantov Digital Creativity