Using 15 year trends to replicate GISTEMP average surface temperature anomalies

At Jo Nova’s Unthreaded on 22/06/14, Philip Shehan posted some GISTEMP temperature trend figures that caused a good deal of controversy.

In 15 year steps they are

1924/39 Trend: 0.142 ±0.148 °C/decade (2σ)

1939/54 Trend: -0.088 ±0.144 °C/decade (2σ)

1954/69 Trend: 0.024 ±0.151 °C/decade (2σ)

1969/84 Trend: 0.165 ±0.162 °C/decade (2σ)

1984/99 Trend: 0.234 ±0.167 °C/decade (2σ)

1999- Trend: 0.099 ±0.138 °C/decade (2σ)

Two issues with the trends are

1. They do not really capture the trends in the data.

2. They slope of the 15 year OLS lines is sensitive to shifting the period by one year – for instance replacing 1999-2013 with 1998-2012.

I replicated Philip Shehan’s data (or a least tried to – he does not use J-D years) on a graph, along with shifting the periods a year backwards. Compare these slopes to the 5 year centred moving average curve in light blue.


  1. Philip Shehan

     /  13/05/2016

    Just came across this but missing the point. It does not matter if the 15 year periods are J to D or if you start a year or more before or after the periods I chose. .

    In fact that is my point. Fifteen year periods are too short to be meaningful in terms of long term trends. The size of the confidence limits indicate that. and at the time of writing the 15 year trends from 1998, when skeptics like to start the pause because it is at the top of an extreme el nino event, which has sod all to do with anthropogenic global warming due to CO2, reduces the magnitude of the trend has become a stistically significant warming trend with about the same magnitude as that for the last 30 years.

    As of April 2016, 18 years and 4 months after the beginning of 1998, the “pause”, which was always a non statistically significant reduction in the warming rate due to natural short term variation in temperature forcings like all the other 15 year variations in my original data has become a statistically significant warming trend of

    0.161 ±0.117 °C/decade.

    The pause was always a statistical phantom, helped along by a cherry picked start date.

    • manicbeancounter

       /  13/05/2016

      In terms of GHG emissions, 1998 is also significant. After a recession in many emerging economies, the rate of emissions growth increased. There should have been no pause if the central estimates of climate sensitivity are correct.
      You also talk about the pause as if it were something of the past. We are at the tail end of another extreme el nino event. It is only in a couple of years that we will know.

      Furthermore, the puase is well-accepted in the peer-reviewed literature as are the large number of excuses. Why should your explanation be the only one?

      • Philip Shehan

         /  14/05/2016

        There is no evidence that 1998 was in any way exceptional in terms of Greenhouse gas emissions. Here is the atmospheric CO2 concentration and Roy Spencer’s UAH satellite temperature data.

        Note that I have stopped the temperature trends from 1998 and 1998 at the end of June 2015, excluding the 2016 el nino spike from the trends.

        I have done this because remarkably, those who want to start the “pause” with the el nino event of 1998 have told me that it is unfair to end with the el nino event of 2016.

        The have also started telling me that as el nino events have sod all to do with CO2 it should not be included in such trends.

        Now I have been pointing out precisely this fact for years now with regard to their insistence that the “pause”, which they say discredits the connection of CO2 concentration with temperature, must start with the el nino event of 1998.

        Further, they say that I must not start a temperature trend one year after the el nino event (1999) because that is a la nina year. That would be unfair or unreasonable. Yes, really.

        But I have played along with these totally ludicrous restrictions and pointed out that their “pause” from 1998 shows a non statistically significant reduction in the warming rate due to natural variations (including the el nino event) which are superimposed on the underlying steady increase in temperature due to the increasing CO2 concentration, evidenced by the CO2 /temperature data since 1979.

        So people offering expanations or “excuses” as watts and you prefer, for the pause are starting with an entirely false premise.

        There was no “pause” to explain or excuse.

        So I do not claim the “pause” was in the past. There is no statistically significant evidence that there ever was a “pause”.

        It is a failure on the part of those who feel they need to explain it by what you term “excuses” to recognise that this is just another short term variation in the long term warming trend due to natural variations which were evident in my original graph breaking the temperature record up into 15 year segments.

        My original graph shown below started backtracking in 15 year segments from 2013, the year I first posted it in response to claims about then alleged 15 year “pause”. The next year I updated the “to” and “from” dates by a year.

        Note that in WFT plots “from” begins at the beginning of the nominated year and “to” ends the selected data set at the beginning of that year. So 1998 to 2013 is 15 years inclusive. You can verify this by changing the first day set “from” and “to” parameters to 1922.99 and 2012.99. There is no change to the plotted data.

        Again, the original graph starting the 15 year period in 1998 was based on the “skeptics” cherry picked start point. But again it is, according to these people, entirely ilegitmate to begin or end the time period at other points in the ENSO cycle which refutes their claims about a “pause.” This is why I put the ” ” around “skeptics”.

        Note that the CO2 temperature data give an empirical value of the equilibrium sensitivity parameter, the temperature rise with doubling of CO2 concentration, of 2.2 C for UAH data from 1979 (when satellite temperature data began), 2.5 C for Gistemp data from 1958 (when Muana Loa CO2 data began) and 2.5 for Gistemp data from 1979.

        These values are within the IPCC range of 1.5 to 4.5 C.

        • manicbeancounter

           /  14/05/2016

          You have given a very long explanation based on your own opinions, and a couple of Wood-for-Trees graphs to back that up.The normal backing people do in these circumstances is to refer to the opinion of others. But you reject the expert opinion of others. In fact, in the “excuses” are opinions similar to what you express.
          With complex and highly variable data, which can be estimated in an infinite number of ways, it is important to demonstrate to demonstrate why your particular explanation is the best one. That can only be done by reference to the other explanations and other sources of data. The alternative is what you are doing – in line with other supporters of climate alarmism – is to vaguely sort the data to fit your view of how the world ought to be.
          Your explanation of the equilibrium sensitivity parameter (which is not referenced) shows that this. They are not the only ones. As Jaime Jessop has shown at cliscep blog, as more actual data comes available, the sensitivity estimates become smaller.

          Kevin Marshall

          • Philip Shehan

             /  14/05/2016

            You have given a very long explanation based on your own opinions, and a couple of Wood-for-Trees graphs to back that up.”

            Yes it is long. That is, comprehensive. So is this one.

            As for the wood for trees graphs. Do you dispute the accuracy of the data? I have seen skeptics use these graphs to back their case.

            “The normal backing people do in these circumstances is to refer to the opinion of others. But you reject the expert opinion of others. In fact, in the “excuses” are opinions similar to what you express.””

            Completely and utterly wrong. A scientist looks at the data and interprets it. (It’s my day job.)

            A scientist never defers to authority. A scientist may in fact think that the established opinions are wrong and set out to show why they are wrong. If successful, you get much more brownie points than if you produce a study that goes along with the crowd. Manuscript referees and thesis examiners will tell you so. Mine did.

            That is the problem Galileo had with the church. In Bertoldt Brecht’s play The life of Galileo there is a scene where Galileo is imploring the churchmen to look through his telescope at the moons of Jupiter. They reply there is no need because the Divine aristotle has explained why they cannot exist.

            The fact is that looking at the data as I have requires no theoretical assumptions . It is a simple straightforward sttement of a striaghtforward mathematical procedure known as a linear regression which “skeptics” use to claim a pause.

            I have not vaguely sorted any data. It is the “skeptics” who have very deliberately been very, very restrictive in choosing data on scientifically illegitimate grounds to back their desired conclusion. But I have been prepared to go along with that restriction.

            The trend and 2 sigma confidence limits which they denmand strt with the le nino event of 1998 but ending 30 June 2015, before the el nino event of 2016, are

            UAH satellite data 0.076 ±0.189 °C/decade

            Gistemp surface data 0.121 ±0.110 °C/decade

            From 1979 the trends are respectively

            0.139 ±0.065 °C/decade

            0.159 ±0.040 °C/decade

            For short term periods (18.5 years here), the confidence limits which define the boundaries between which there is a 95% chance that the true value lies are quit large.

            This is because for short term periods, the “noise”, the fluctuations in the data away from the trend line are large compared to the signal, the vertical rise in the trend line, over that period.

            For the GIstemop data, the entire 95% range is above zero (the trend value is larger than the confidence limits) so the trend actually shows statistically significant warming.

            For the longer 38.5 year period, the signal to noise ratio is higher, and the confidence limits smaller. Both data sets show statistcally significant warming.

            Moreover, as the error margins for the short and longer term periods overlap, there is no statistically significant difference between the data.

            That is, the data supports the null hypothesis. That any deviation between the long and short term data can be attributed to chance.

            So, even obeying their ludicrously restrictive cherry picking of the start and finish dates in order to get the ENSO cycle, which has nothing to do with CO2 concentration, to back their claim that temperature has not kept pace with CO2 concentration therefore the basis of AGW is “proven” to be false, cannot hide the mathematical fact that the data shows there is no statistically significant variation from the long term trend value.

            No “pause” or “hiatus” (which “skeptics” used to say meant “no warming”), but no even a statistically significant reduction in the warming rate.

            If I am permitted to remove the scientifically untenable restriction that it’s fine to have an el nino event at the start of a data set but not at the end, we have statstically significant warming trend of from 1998 to April 2016 (nothing vague about that time period either)

            Trend: 0.161 ±0.117 °C/decade.

            Now if I reaaly want to upset the skeptics, I could entirely justifiably (by their logic) reverse their restriction and start after an el nino event and end by including one;

            UAH data from 1999 to April 2016:

            Trend: 0.185 ±0.179 °C/decade

            No pause. A large statistically significant warming trend.

            NOT FAIR cry the “skeptics”.

            Simple mathematical fact entirely independent of any theory behind the data.

            The data could represent stock market prices, salmon population in the north pacific or anything else.

            The mathematical conclusion stands alone.

            And yes as more data becomes available and the estimates become more accurate, widely varying estimates of the sensitivity parameter (or anything else) converge on the true value.

            Notice how well the ECS calculated from the CO2 /temperature data here fits with the convergence point.

          • manicbeancounter

             /  15/05/2016

            I wrote this post three years ago. Since then I have move on, and the climate “debate has as well”.

            You say with reference to the excuses for the pause.

            A scientist looks at the data and interprets it. (It’s my day job.)
            If you look at the type of complex, adjusted, homogenized (processed) and aggregated data that are in the surface temperature records you will understand that with each processing you will get a different answer. This can be quite sensitive to changes in adjustment assumptions. You end up with any number of slightly different explanations. It becomes like economics, which I studied. There you only can say so much about the data results.
            Greater precision can only come about by properly understanding how the data is compiled, then looking at how this can be improved. This I did last year in a series of posts. The two concluding posts are Defining “Temperature Homogenisation” and Climatic Temperature Variations.

          • Philip Shehan

             /  15/05/2016

            You may have written the initial post three years ago, but unless I am mistaken, you are still claiming there is a pause. “Skeptics” have clearly failed to “move on” as far as that goes.

            I fact I am still having this argument over at Mr Bolt’s blog, and came across your post here while chasing up some arguments on this topic.

            But if you now wish to aver that with another 3 years of data it is apparent that there actually was no pause, or that there is no longer a pause, go right ahead.

            And again I must point out that that I am not making “excuses” for the “pause” I am saying that the mathematics does not support the contention that there ever was one. So no explanations or “excuses” are required.

            You want to claim that my analysis of the temperature data is flawed because the temperature data is flawed. But skeptics base their claim that there was a pause on the same temperature data.

            This is as logically tortured as the argument that you must start a temperature data set with an el nino event and not end with one, but not vice versa.

            Cherry picking particular data sets on the basis that some are adjusted and therefore unreliable is nothing more than an attempted evasion of unwelcome facts.

            ALL the data sets are adjusted, including the satellite data. Those I have used here are RSS version 3.3 Tropospheric Lower Temperature and UAH version 5.6 Tropospheric Lower Temperature.

            Even Carl Mears of RSS says that satellite data is no more reliable than surface data and that no single data set should be viewed in isolation.

            But the fact is that the differences in trends given by the seven data sets I used are negligible compared to the error margins, which means they are all statistically equivalent.

            Meaning that whatever data set you choose, and that includes RSS data only, there is no statistical difference between the period from 1998 to 30 June 2015, (the period obeying the skeptics demands), and the trend for the period from 1979.

            So there is no statistically significant reduction in the warming trend, let alone a pause.

%d bloggers like this: