I start with a bit of housekeeping. The website began on 16 July 2012, with a post explaining why I was going down this path. It drew no comments, but the next one, on the carbon tax, drew three. Since then there have been just over 700 posts, which is a lot of writing — I’d estimate 750,000 words. And there have been nearly 6,000 comments. In all, 48,000 unique individuals have come to the site, for 130,000 sessions, with 320,000 page-views — on average, each visitor reads 2.5 pages, or two-and-a-half posts. Each post seems to attract about one-third new visitors, and some stay on, because the ‘traffic’ just keeps growing.

Last year most posts drew about 200 to 250 readers on day one, and then attracted between 1000 and 1500 readers in the first month. But old posts keep being re-visited. Who are the visitors? Three-quarters of them are under 45, according to Google Analytics. Some 55 per cent are men. The great majority, about 85 per cent, live in Australia, most of the rest in the USA and UK. My thanks to you all, and especially to those who comment. It is a civilised website, and I have not felt the need to moderate any post to the point of elimination. It is perhaps a sign of the website’s growing traffic that it now attracts a lot of spam — and that causes some extra work for me in examining each piece and coming to a decision. Much of the spam turns out to come from commercial sites offering watches, shoes, Viagra, gambling, and sex of various kinds.

It follows from all this that managing the website is a lot of work, and I was warned that this would be the case when I began. I don’t need donations and I don’t need a helper, but I do need to restrict what I do because my other writing is pressing. So for this year I will write one post for each Monday, but not any more unless something turns up about which I want to comment quickly. There will be elections this year both for the Commonwealth and for the ACT, where I live, and some of my writing will be about those two events.

I don’t want to write so much about ‘climate change’, because I feel I have said everything I want to say about that issue, which doesn’t seem to alter much. The Paris Agreement has disappeared from sight, and while extreme weather events have been occurring in the UK and the USA, and we have had fires here, the clamour that these events must be caused by greenhouse gas emissions seems to have subsided. More about extreme weather in a moment. I intend to develop a place on the site for my position on ‘climate change’, so that I don’t have to repeat myself. I’m not quite sure how this will work, but stay tuned.

At the time of writing only Roy Spencer has published a piece about whether or not 2015 was the warmest year ever, or since some kind of previous record was established. I understand that the land and sea data on temperatures will be released about January 20th. The UAH satellite data (along with those from RSS) show 2015 to have been the third warmest year, after 1998 and 2010. But there were certainly many claims that the year would ring that particular gong. The main cause for the warmth is hotter sea surface temperatures, themselves the result of the prevailing el Nino in the Pacific, which may have peaked. Floods and storm in the USA and the UK have occurred, and have produced familiar claims that these extreme weather events are examples of ‘climate change’ providing additional reasons why the climate models are right, and we must get rid of fossil fuels.

The difficulty with this sort of claim is that it is easy to point to any number of previous storms and floods that could not have been caused in this way. In October 1987 I was in London to give a lecture at the Institute for Commonwealth Studies. I stayed at the Crest Hotel across Russell Square, and woke up to hear a steady loud banging from outside. I got up and went to the window. The lid of a large garbage hopper was rising and falling, making a dreadful noise. Obviously wind was causing it, and I could hear wind shrieking. Eventually I went back to bed. In the morning I discovered that the power was off, and there would be only a cold breakfast. A great storm — a cyclone! — had descended on the UK.

It was not until later in the day that I understood what had happened across the south-east of the country. But London had experienced its first total blackout since 1945, the railway system was not running in the south east, a score or more of people had been killed, and Russell Square, which I had to negotiate to go the Institute, was an astonishing mess. Great trees that had been the glory of the gardens had been uprooted — they were 170 years old. Caused by ‘climate change’? No one said so at the time. It wasn’t an issue.

Had anyone argued in that way, one could point to Daniel Defoe in 1703, who was a spectator to another cyclonic event, in which violent winds killed more than 8,000 people.   Defoe had given up counting the fallen trees in Kent when he reached 17,000. He wrote a book about it, The Storm, whose cover looks like this:


The London Telegraph, alarmed at the beginning of this month that so few people seemed to have any idea that extreme weather had ever happened in the UK before, provided a helpful list of earlier examples, with illustrations, which you can consult here.

It’s much the same in Australia, with weather ‘records’ being referred to in the media that have no substance at all, when you look at the history. Brisbane had a ‘horror’ flood a few years ago, and our then leader, Ms Gillard, instituted a flood levy to provide for infrastructure rebuilding and some disaster relief. As it happens, the Bureau of Meteorology maintains a website on the Brisbane River floods. The Bureau lists numerous floods, which it categorises as Major, Moderate and Minor. Of the nine Major floods, the recent Brisbane one was number seven in height, and nowhere near the height of the really big ones.

As with floods, so with extreme temperatures. The Bureau also keeps a list of extreme temperatures, as you might expect, and the years of these records are all over the place. The hottest was in South Australia’s Oodnadatta, in 1960. Then comes WA in 1998, NSW in 1939, Queensland in 1972, Victoria in 2009, the Northern Territory in 1960 and Tasmania in 1960. For completeness I add the ACT, the hottest temperature in which occurred in 1968. Do you see a trend in all of this? I don’t. Of course, the Bureau would like us to look at temperature anomalies, but even there I don’t see much of a rise over nearly a century.


The evidence suggests that Australia is plugging along as it has done for a long time now, and if it turns out that 2015 was indeed the hottest year since whenever, I don’t think that should perturb anyone, unless of course they are Climate Botherers.

Thought for 2016: John Cleese, yes, that one, has apparently tweeted: ‘I would like 2016 to be the year when people remembered that science is a method of investigation,and NOT a belief system.’ I’ll drink to that.


[Later note. I have changed the sentence about Julia Gillard’s flood levy to make clear that most of it was for infrastructure repair, though some disaster relief was included.]

Join the discussion 94 Comments

  • whyisitso says:

    Don, to me, the graph in your post shows a very distinct upward trend. Perhaps you could educate a non-numerate guy like myself on your quote: “Do you see a trend in all of this? I don’t.”

    • donaitkin says:


      I said ‘I don’t see much of a rise’. I could have added that you can see sharp rises and sharp falls, and that the average wobbles around, too, in its gentle rise over a century. Is that better?

      • whyisitso says:

        I misread your paragraph. On my first read I thought you were referring to the graph (which does show a pronounced upward trend). But of course you were referring to the preceding text about the hottest temperatures in various places, not the graph. My apologies.

        • Gerard says:

          Note the BMO chart starts at 1910, ignoring data from earlier years which were warmer. See below:

          ‘It is as if history is being erased. For all that we hear about recent record-breaking climate extremes, records that are equally extreme, and sometimes even more so, are ignored.

          In January 1896 a savage blast “like a furnace” stretched across Australia from east to west and lasted for weeks. The death toll reached 437 people in the eastern states. Newspaper reports showed that in Bourke the heat approached 120°F (48.9°C) on three days (1)(2)(3). The maximumun at or above 102 degrees F (38.9°C) for 24 days straight.

          By Tuesday Jan 14, people were reported falling dead in the streets. Unable to sleep, people in Brewarrina walked the streets at night for hours, the thermometer recording 109F at midnight. Overnight, the temperature did not fall below 103°F. On Jan 18 in Wilcannia, five deaths were recorded in one day, the hospitals were overcrowded and reports said that “more deaths are hourly expected”. By January 24, in Bourke, many businesses had shut down (almost everything bar the hotels). Panic stricken Australians were fleeing to the hills in climate refugee trains. As reported at the time, the government felt the situation was so serious that to save lives and ease the suffering of its citizens they added cheaper train services: “The Commissioner of Railways promised a deputation of members of Parliament to run a special train every Friday at holiday excursion rates for the next month to enable settlers resident in the Western part of the colony to reach the mountains to escape the great heat prevailing.” (JoNova Nov 6 2012) Should read full article.

    • chrisl says:

      Whyisitso Look up chartsmanship and look at the scale of the y axis

  • Alan Gould says:

    Good on John Cleese. I look forward to the ‘Post’ year, Don.

    • Peter Kemmis says:

      Yes, it is encouraging when a notable figure in the arts steps forward to call a spade a spade. Perhaps the sycophancy towards the AGW meme that we have seen from so many of today’s leaders in the arts, is an aberration. For so often it has been the artists who have stood up, and cried “hey, what’s going on here?”

      • dlb says:

        Clive James is another climate sceptic from the arts. Probably much to the chagrin of the ABC who have generally regarded him as one of their darlings.

  • Congratulations Don, you are doing well for a relative newbie in the blogosphere.

    Rafe (the grand old man of the blogosphere) Champion

    That designation came from Ken Parish of Club Troppo, in happier times before I fell foul of Ken and his Troppo colleagues on the topic of climate change.


    • donaitkin says:

      I went back to your link. It was five years ago, but what astonishing vitriol! Is it still the same? I occasionally go there because someone sends me a link, but I’m not a frequenter.
      And thank you for your nice words.

      • whyisitso says:

        Ken Parish was one of the original bloggers on the blogosphere in the early 2000s. I used to comment on his blog using my real name, but I eventually realised how dangerous that was.

        I do recall Rafe’s post that in my view was a reasonable short post, but was obviously too politically incorrect for Ken and Nick Gruen.

        One commenter I was extremely disappointed with was Geoff Honnor, who I’d always thought of as a voice of sanity – not here unfortunately.

        Ken Parish had been a fairly thoughtful blogger in the early days until he surrendered his blog to far leftists Mark Barnisch (Lavartis) and later, Nick Gruen. The blog now inhabits the far left segment of the blogosphere, intolerant of any post or comment that doesn’t go along with their distorted view of the world.

        I made a comment on the blog a year or two ago and have since been banned. i don’t even recall what I said that was so outrageous.

  • chrisl says:

    Hello Don, Just out of interest, do your blog posts get more readers than your other writing? The world certainly needs more reasoned writing than what is thrown up by a twitter storm that eventually ends up in the main stream media

    • donaitkin says:


      It’s impossible to say. I used to write a weekly newspaper column, and people would tell me that they read my essays, but what proportion of the weekly buyers actually read my piece? I assume the proportion was reasonable because otherwise the editors(s) would have dropped me. But if they did surveys they never told me the outcome!


    • Mark says:

      ” I used to write a weekly newspaper column”

      Yes, as I recall the inside backpage of the National Times. Perhaps oen day when you’re in a reflective mood you might write about the demise of quality journalism that was the hallmark of the NT. Or perhaps you still see it around. Or perhaps my glasses are too rosy.

      Still I’d be interested in your thoughts.

      • donaitkin says:


        It’s mostly the decline of advertising for the newspapers, and thus a much smaller staff than used to be the case. Few papers have the reserve capacity to allow a reporter or two to follow up a good but slow story.

        And in some ways the blogosphere has taken over. A lot of stories start there, and the papers then pick them up.

        I don’t think it is a conspiracy, or an aversion to the truth. Just papers trying to find out how to keep heads above water.

  • David says:

    Do you see a trend in all of this?”

    Yes. 7 of 8 of the all time daily maximums occurred after 1950, while 1 of 8 occurred before 1950. That is an increasing temperature trend.

    • donaitkin says:


      • David says:

        “In 2013, the UN’s top climate official, Christiana Figueres, linked bushfires in Australia to climate change. Abbott called such claims “complete hogwash” and said drawing links between broken records and climate change was a sign of desperation.”

        Abbott makes the claim,

        “The thing is that at some point in the future, every record will be broken, but that doesn’t prove anything about climate change. It just proves that the longer the period of time, the more possibility of extreme events.”

        Sophie Lewis in the journal of Weather ad climate Extremes pulls Abbott’s comments apart.

        “The first way to understand Abbott’s claim is that in any system, the longer you wait, the more often you will see records fall. But Lewis points out that the exact opposite is true. In a system without any sort of trend, such as a random string of temperatures, the first one will be a record-breaker, by default. The second one will have a 50% chance of being a record-breaker. The third has a one in three chance of being a record breaker … and so on. In a very long temperature series, you should see very few records being broken, and they will break less often over time.”


        And in the set of temperature records you provide we see the opposite effect, with records being broken more frequently with he passing of time , not less frequently as one would expect if there was no temperature trends.

        • JimboR says:

          Yes, Abbott’s comments are a great example of what happens when you rely on common sense instead of maths to solve scientific problems. Kinda’ sounds reasonable until you stop to think about it.

        • donaitkin says:

          It’s hard to be sure which bit you are attacking. I would agree with the former PM that Ms F is talking hogwash in linking bush fires to climate change, and I would feel that the same is true when people link the current blizzard in NE USA to climate change.

          But in typical fashion, Mr Abbott got the next bit wrong, at least I think he did. In a period when temperatures have been rising for 150 years, it is likely that from time to time records for high temperatures will be broken. That’s not what he said, but it is possibly what he wanted to say. I agree that temperatures have been rising, but I have not been persuaded that the cause is carbon dioxide.

          David, you possibly prefer your role as gadfly, but if you have something substantial to argue, you can have a thousand words to do it here, as long as it’s about global warming or climate change….

          • JimboR says:

            “In a period when temperatures have been rising for 150 years, it is likely that from time to time records for high temperatures will be broken. ”

            Do you think it’s also likely then that the frequency of high or extreme fire danger days will also increase?

    • Gerard says:

      Adjusted or unadjusted (homogenised) data?

  • Anthony says:

    Don, what a great fortune it is for me to have stumbled across your website! It is refreshing and encouraging to find others of a likewise mind. I look forward to exploring your website and your writings with interest.

  • Michael Cunningham aka Faustino says:

    Hi, Don, just read this and the previous post. I’ve been a regular at Judith Curry’s blog since its inception, but, like you, have just stepped back from the alleged CAGW issue, I’ve given it a lot of time over many years (I was briefed by the IPCC’s Chief Scientist in 1989 or ’90, I have nothing new to say and it’s time for a change of priorities.

    Thanks for your continuing efforts and for a very civil blog, I hope that more pressing issues will now benefit from your attention.

  • dlb says:

    re the 2011 Brisbane floods, I have heard it said by the “authorities” that had Wivenhoe Dam not been in existence then the Brisbane flooding would have been worse than the record flood of 1893.

    Still this in no way proves that bad weather is increasing, more to the point it makes Flannery a fool for saying the dams will be unlikely to fill again.

  • dasher says:

    Thanks Don. I value your opinion and I will be interested to hear what you say about climate change. Having read a great deal about the Paris conference, politicians babbling (including the leader of the free world) and journalists picking and choosing but never really informing (on both sides). Judith Curry nailed it before Christmas. Your turn.



  • David says:

    Interesting link

    • donaitkin says:

      Thanks for the link. Yes, it is an interesting little video. One of the commenters seemed to me to get the message: the satellite data are consistent with the balloon data, and over the whole 35 years or so. So if the satellite data are wrong, how is it that the ballon data are wrong too, given that they are completely independent?

      Yes, you should look at all the data, and very broadly they show the same trend — a noticeable increase in temperatures from about 1975 to 1995, and a slower rise, even a stasis, thereafter. How steep and how slow depend on which dataset you want to use. In all of them the error bounds are probably larger than the increase.

      • David says:

        Towards the end of last year Bobo was discussing the merits or otherwise of using a Beta version of one of Dr Spencer’s “retrieval algorithms”, site unseen. Having watched this video I can see why Bobo was skeptical.

        • JimboR says:

          Oh dear, it seems Don’s worst fears have been realised. The data-fiddlers have got their hands on the satellite data too. Where to now?

    • donaitkin says:

      I made no comment on this video, if only because I find Michael Mann objectionable as well as deeply untrustworthy. However, I did watch the video, and in return I suggest you read Christopher Monckton on what is wrong with it. Now you probably don’t rust him, but I think you need to read what he has said.


  • David says:

    Don you have made statements like this many times.

    “In all of them the error bounds are probably larger than the increase.”

    I presume this statement is meant to cast some doubt on the notion of a temperature trend. What statistical test are you referring to?

    If you regress temperature on time the coefficient on time will be statistically significant. The trend is positive.

    I think when correctly interpreted, both balloon data and satellite data, indicate temperature is increasing.

    • JimboR says:

      You’re a patient man David. You’ve attempted to explain this countless times, but still Don persists with his inference that you cannot extract precise information from a very noisy signal. When pressed, he’ll say he never made that claim, but continues to liberally spread FUD with comments like the one you quoted.

      Perhaps the promised “place on the site for my position on ‘climate change’” will clarify all these twists and turns into a logical, rational, scientific position.

    • donaitkin says:

      My statement is intended to cast doubt on the apparent precision of the trend. As I have also explained many times, there are two sorts of error, and the chief one, for land and sea data, lies in the number and placing of the recording devices — and, in the past, the likely error in recording what was found (buckets for sea temperatures etc).

      As I have also said many times, I accept that the planet has warmed over the 20th century. But for me statements about ‘the warmest year ever’ are pretty silly.

      • JimboR says:

        “My statement is intended to cast doubt on the apparent precision of the trend.”

        And the counter point is that you can extract extremely precise data from extremely noisy input signals. To deny that is to deny countless scientific and engineering fields.

        • donaitkin says:

          Perhaps you could show us, with the data on temperature that we have.

          • JimboR says:

            I’m not going to attempt to jam an entire undergraduate course into a reply. I’m sure I wouldn’t do it justice, but I am happy to recommend one.

            The good news is there are many diverse fields that all use much the same techniques, so there are plenty to choose from: statistics, signal processing, machine learning to name just a few. My favourite is probably Caltech’s CS156:


            In just 18 free one hour lectures you’ll have all the tools you’ll need. The maths is quite manageable provided you’re not scared of a bit of matrix algebra and combinatorial algebra.

          • JimboR says:

            I thought we were discussing error tolerances, hypothesis testing, probabilities, likelihoods and your intention to “cast doubt on the apparent precision of the trend”.

            I’ll repeat David’s question:

            What statistical test are you referring to?

          • donaitkin says:

            Jimbo, you are over-complicating things. I was asking you to show us how you could detect the signal from the noise, given that no one else seems to be able to it.

      • David says:


        If you say you accept the plant has warmed then you cant then say

        “In all of them the error bounds are probably larger than the increase.”

        To me the meaning of those two statements are at cross purposes.

        • donaitkin says:

          Not necessarily: ‘in all of them’ applies to the datasets, each of which shows movements in the anomaly by day, month or year. Each time there is a movement , it seems likely to me that the errors are larger than the shift. If the movement is over a century, then no. I would accept that there has been a movement of consequence. Whether it was 0.7 degrees C, 0.8 degrees C or 0.6 degrees C is another matter altogether.

          • David says:

            “In all of them the error bounds are probably larger than the increase.”

            And this

            “Each time there is a movement , it seems likely to me that the errors are larger than the shift.” ??

            What do you mean by error bounds?

            Here is a standard formula for a confidence interval.

            X = +/- z * (S/?n)

            X = sample mean
            z = z score
            S = Sample standard deviation
            n = sample size

            Two things you should focus on.

            1. As n increases, X (aka error bounds) decreases.

            2. There is no slope term in there.

            So I don’t understand why you keep trying to compare “error bars” to a “shift”.

            If you can show me an example of where your statistical methods have been used that would be really helpful.

  • JimboR says:

    Actually the proceeds of the flood levy went to rebuilding government (all levels) owned infrastructure, not to “all the people who had built on the flood plain and had no insurance”.

    • donaitkin says:

      Yes, my error. People affected were helped through a Disaster Relief Fund that already existed. The levy was for roads, bridges and the like.

      • JimboR says:

        And most of the Disaster Relief Fund came from public donations. So we all got to choose whether or not we supported “all the people who had built on the flood plain and had no insurance”, and we all had to chip in to rebuild public infrastructure (at least those of us earning more than $50K).

      • JimboR says:

        Easily edited though right? Or are you letting the error stand as it suits the ideology?

  • donaitkin says:

    ‘Early figures suggest the direct total cost of the floods to the federal government will be $5.6 billion, including rebuilding and disaster assistance payments.
    Rebuilding in Queensland is estimated to cost $3.9 billion, while the damage bill for other flood-hit areas was put at $1 billion.’

    SMH January 17, 2011

  • donaitkin says:

    The nesting here has got out of hand.

    For Jimbo and David above: I am not talking about statistical error and the tests relevant to it. I am talking about measurement error. Sea surface data, until very recently relied on a variety of data produced from a variety of devices, including leather buckets and engine intake manifolds. Datasets have used these data without any reference to the uncertainties involved. It passes belief, at least to an agnostic like me, that indicated changes of tenths of a degree C are meaningful when the measurement errors must be greater than this.

    There is a useful essay on uncertainty at http://fabiusmaximus.com/2016/01/15/measuring-error-in-ocean-warming-93036/

    Worth a read.

    • David says:

      Measurement error is sub-set of statistical error.

      I am not familiar with the details, but working with your example these “leather buckets and engine intake manifolds” these imperfections cause some estimates to overshoot the “real” temperature and others to undershoot.

      Statistics makes this grand assumption that the expected value of the error term is equal to zero. And the larger the sample the more certain we can be that the error term will be zero. (Apologies to all the statisticians out there if I have mangled this explanation).

      This is what JimboR is saying. Even though there maybe considerable volatility in the observed data, increasing the sample size allows us to observe underlying trends.

      • donaitkin says:

        ‘Measurement error is sub-set of statistical error.’ Who provided this odd definition? There is sampling error and measurement error. They are distinct.

        • David says:

          This confusion could be resolved if you would nominate one statistical text, which we could all refer to.

          As I understand it statisticians deal with two fundamental sources of uncertainty.

          The first is “error” or “random error” or “white noise”. The expected value is zero.

          The second is “systematic error” or “bias”

          I can honestly say, I find the way you discuss statistical evidence and statistics to be very confusing.

          • JimboR says:

            Amen to that! You’d think somebody who “once taught this stuff” would want to use the tools to prove his point, or disprove someone else’s claim.

            Don, here’s a really simple experiment to try. Pick a linear formula, I’ll use the same one I picked for my example above:

            T = 0.7t + 23, where T is Temperature (C) and t is time (mins)

            then introduce a massive measurement error. Let’s say your thermometer is only good to nearest 5C, so your readings are good to +/- 2.5C, and your data now looks like:

            Time true temp rounded temp

            0 23.0 25.0
            1 23.7 25.0
            2 24.4 25.0
            3 25.1 25.0
            4 25.8 25.0
            5 26.5 25.0
            6 27.2 25.0
            7 27.9 30.0
            8 28.6 30.0

            55 61.5 60.0
            56 62.2 60.0
            57 62.9 65.0
            58 63.6 65.0
            59 64.3 65.0
            60 65.0 65.0

            Now do a linear regression on the rounded temp Vs time. You’ll get a slope of 0.69566 with a p-value of 5.4E-57.

            Do you at least accept that given the right type of measurement error, it is possible (given enough data points) to get much smaller “error bars” on the trend slope than you had on any of the datapoints going into the experiment?

          • donaitkin says:


            I am not here to engage in a quiz show. If you want to show how the SST data are reliable, just go and do it. I think it is not just very hard, but virtually impossible.

            But have a go. Don’t provide me with little classroom exercises, but go to the big world. Get the data for 1950 for the oceans in the Southern Hemisphere. You can find them. You will note that you don’t have either a population, in the true sense, or a random sample. You will note that vast areas of the oceans have no data at all. You will note that the measurements that exist come through a variety of mechanisms and were obtained on different dates and at different times of the day, in some cases. You will note that the great majority of the data come from the most favoured shipping routes.

            Now you will have to make some assumptions if you are going to talk about SST. You can in-fill data from known points (for how far, and with what confidence?). You might assume that a particular datum applies for a given time, based on work done elsewhere. You might assume that it doesn’t really matter about the difference between engine inlet manifolds and bucket measurements, or that there is no real difference between leather buckets and steel buckets. But each time you do that you are introducing measurement error.

            How are you going to present what you have found? If you had one Argo buoy in every square km of sea, then you might be able to talk about the characteristics of a population, in which case you simply say what the outcome is, using ordinary statistics.

            You don’t have a random sample, and by in-filling you still don’t have a random sample, because the only data you really have are the data you started with. Will you talk about sampling error? What will you say?

            And what will you say about the measurement errors involved in collecting the data? What would your graph look like to a reader, if you did all this for the decade of the 1960s?

            I’ve done a lot of reading here, and I simply don’t think we know very much at all about SST in 1950. About 2015? Yes, we know a lot more.

            But the whole point of all of this is the trend over time. Many, many others beside myself have come to the conclusion that the data from the past cannot be used for that purpose.

            But you think they can. So, show me and any other readers who are somewhat sceptical.

          • JimboR says:

            And I’m not here to convince you the SST data is reliable. I’m simply asking you to justify this statement:

            “It passes belief, at least to an agnostic like me, that indicated changes of tenths of a degree C are meaningful when the measurement errors must be greater than this.”

          • JimboR says:

            When I’m assessing whether I’ve come to the right place to learn about something (SST reliability say), I tend to assess how the author argues his case. When I see statements that I know to be wrong in my own field, it doesn’t inspire me with confidence.

            In general (not related to SST say), do you still stand by your claim I quoted above, and if so why… preferably using the maths you once taught.

    • JimboR says:

      “It passes belief, at least to an agnostic like me, that indicated changes of tenths of a degree C are meaningful when the measurement errors must be greater than this.”

      Then once again I encourage you to take a course on the topic. All physical measurements have errors and the techniques to deal with them are very well established and used in many diverse fields.

      • donaitkin says:

        Yes, yes… You keep saying this. I once taught this stuff. What I am saying is that it is impossible to determine a persuasive estimate for the error in, say, SST measurements when (i) the measurements are taken from different parts of different oceans, (ii) by different devices, and (iii) at different times.

        In fact, no error estimate is usually given when these data are put up in graphs. It is for this reason (and others, like coverage) that I prefer the satellite data, which do have their own problems, though these have been outlined and dealt with.

        • JimboR says:

          “I once taught this stuff.”

          Which text did you use?

        • bobo says:

          “it is impossible to determine a persuasive estimate for the error in, say, SST measurements when (i) the measurements are taken from different parts of different oceans, (ii) by different devices, and (iii) at different times.”

          Don, why is it impossible?

          • donaitkin says:

            Too little real knowledge.

          • bobo says:

            “Too little real knowledge”
            Very vague Don, you haven’t given any justification at all apart from restating your claim.

            It’s a superficial answer that is probably based on gut feeling, on “common sense”.

            You’re making a very substantial, nontrivial claim that you seem unable to back up with evidence or reasoning.

    • JimboR says:

      Don, one of your most often cited issues is the precision with which the trend line slope is quoted Vs the precision of the original measurements. For example:

      “It passes belief, at least to an agnostic like me, that indicated changes of tenths of a degree C are meaningful when the measurement errors must be greater than this.”


      “My statement is intended to cast doubt on the apparent precision of the trend.”

      Imagine a high school physics class where students have to observe the temperature change during some process. To keep things simple, let’s assume it’s uncontroversial that in this experiment Temp Vs Time is expected to be linear. Let’s also assume they have extremely accurate crystals with which to measure time down to the nsec but their recently calibrated thermometers only read to the nearest degree C. So in this case errors in Time are minuscule compared to errors in Temp. They have 60 minutes to complete the experiment.

      Student 1 is smart enough to realise you only need two points to nail a line and so takes two measurements:
      t=0 mins, T = 23C
      t=60mins, T = 66C
      He calculates the line as:
      T=0.717t + 23
      but he realises his readings could have meant 22.5C to 66.5C for the steepest results and 23.5C to 65.5C for the shallowest result, so his quoted slope is 0.717 +/- 0.17

      Student 2 is even smarter and realises any two points will nail a line so he takes two measurements:
      t=0mins, T = 23C
      t=5mins, T = 27C
      and takes an early lunch. He calculates the line as:
      T=0.8t + 23
      His steepest result runs from 22.5C to 27.5C and his shallowest 23.5C to 26.5C, so his quoted slope is 0.8 +/ 0.2

      The keenest student in the class takes a reading every minute, and fits a regression line through all 61 points:
      t=0mins, T = 23C
      t=1mins, T = 24C
      t=2mins, T = 24C
      t=3mins, T = 25C

      t=60mins, T = 66C

      Even if you prefer maths-by-intuition rather than algebra, surely you’d concede that these three students get completely different precision on their slope lines. And even the slackest student who went to lunch after 5 minutes, has much better precision on his slope than he did on the temperature readings. All students had exactly the same equipment and datapoint precision.

      I’ll leave it as a research project for you as to what precision the third student can claim, but you’ll need to make up a full set of readings.

      • donaitkin says:

        See below my long response to bobo.

        • JimboR says:

          But specifically… do you accept that, given enough samples, the precision on the slope of the trend line can be way better than the precision of any of the datapoints that went in?

          If YES, why do you keep saying things like “that indicated changes of tenths of a degree C are meaningful when the measurement errors must be greater than this.”

          If NO, why not? Preferably in mathematical terms.

          • donaitkin says:

            Jimbo, you are confusing sampling error and measurement error. They are distinct.

            Sampling erors, to add a bit, are relevant where what we have is a random sample. For land and SST we don’t have anything like that. We have a miscellany of data sites, lots in the USA few in east Asia, central Australia etc. Nonetheless they are treated as though they were indeed a random sample. But they’re not. This is why (see below again) one has to deal with all this stuff in the most general way. Yes, there is likely to have been some warming (glaciers receding much of the time), but what occurred varied across the globe and over time. Don’t make too much of the numbers…

          • JimboR says:

            “Jimbo, you are confusing sampling error and measurement error.”

            The errors the students encountered above were measurement errors, not sampling errors. They needed higher resolution than their lab equipment provided them. They solved that problem with oversampling (well, at least the third student did).

            This is a very common technique. Indeed noise can even assist in this process. It’s common to add white noise into a signal, to make the oversampling more effective aka dithering the signal.

  • Nicole Parton Fisher says:

    As always, a very interesting column, Don. I take it that your broad point is that weather is too often conflated with climate change. I agree with that position. Historically, climate change has not occurred in decades or in centuries, but gradually, over millennia.

    Statistics are an invaluable tool for those who understand and use them without bias. I am, unfortunately, innumerate. But let me say this.

    If I eat five cupcakes in one day, I’ll gain a little weight that I’ll probably lose by hitting the gym. If I keep eating a daily five cupcakes over a long period of time, it will be harder for my body to realize the gym’s benefits.

    When the effect of all those cupcakes gradually surpasses my desire and ability to make my way to the gym, my body will have experienced a visible change in a relatively short period of time (Trust me: I speak from experience on that). In my admittedly unscientific opinion, the world has been eating too many cupcakes.

    Nicole Parton Fisher, Canada

  • margaret says:

    Hear hear …

  • bobo says:

    You’ve written
    “I intend to develop a place on the site for my position on ‘climate change’, so that I don’t have to repeat myself. I’m not quite sure how this will work, but stay tuned.”

    I would be interested to see your stance on global warming clearly stated without internal contradictions, and criteria with which you reject data sets, models and theories, because there are very few data sets (and models) that you feel comfortable with, and you have stated that you are not “interested in theories”. The BOM data that you use and the satellite data that you quote is processed using models based on theories, yet you use these.

    Why do you accept the BOM temperature data despite your scepticism towards land temperature data, and the scepticism towards these sets by someone you trust, Jennifer Marohassy?

    Moreover, which UAH data set do you think is more correct? Is it the officially published set by UAH or the unverified beta set published on on Roy Spencer’s website? A bit of analysis of the two follows:

    The ranking of annual temperatures Roy Spencer publishes on his personal website using his personal unverified beta dataset is:
    1998 +0.48C
    2010 +0.34C
    2015 +0.27C

    I also checked the official UAH data at http://nsstc.uah.edu/climate/
    On that page you see a link to a “December 2015 Global temperature report”, written by Roy Spencer and John Christy. It includes a list of the warmest 5 years:

    1998 +0.48C
    2010 +0.34C
    2015 +0.27C
    2002 +0.21C
    2005 +0.20C

    At this point I want to stress that Spencer’s beta data set is not officially published by UAH and is very different to the data that can be downloaded from the UAH page I’ve linked, the same page for accessing the December 2015 report based on Spencer’s beta data set.

    Here are the averages for the years above calculated using UAH official monthly data:
    1998 +0.42C
    2010 +0.40C
    2015 +0.36C
    2002 +0.22C
    2005 +0.26C

    So 2015 is also the 3rd warmest year according to the official data, but the anomalies are very different and the two data sets disagree on the rankings of 2002 and 2006. Quite obviously the December 2015 report on the UAH site is not based on UAH official data. Spencer has so far refused to release the algorithm for generating his beta data for open scrutiny.

    Don, I’d be interested to know which of the two data sets you accept and which you reject, and why.

    • donaitkin says:

      Being asked to deal with propositions like this is precisely why I want to set out somewhere a summary of my position on global warming/’climate change’. I’ve said many times what I’ve said above, and explained why.

      In brief, (i) the data simply do not allow accurate, precise accounts of global temperature, for all sorts of reasons, and global temperature is a construct that has no real application to any human being; (ii) but to dismiss it all is to leave the debate; (iii) so one has to compromise if one wants to take part in the debate; (iv) mine is to accept that the world has probably warmed over the 20th century, in an irregular fashion; (v) technological improvements mean that the most recent measurements of temperature are more reliable than those of the past; (vi) the Argo buoys and satellite and radiosonde balloons are to be preferred to the land and (especially) SST measurements for reasons of coverage alone; (vii) the science of ‘climate change’ is not settled at all; (viii) it is the mood of the electorate, as discerned by governments seeking re-election, that keeps the structure and funding of climate science going in its present direction, so that what governments and their supporters present, most of the time, is ‘policy-based evidence’, not disinterested argument; and (ix) I see my task, given my own background and experience, as drawing attention to the argument and evidence brought to bear on the issue, which many political leaders, like the past Rudd and the present Obama, have defined as the most important challenge facing humanity. I think that they are quite wrong.

      Of the 700 or so posts I have written, two thirds are about this issue in one way or another, and the questions you raise have been dealt with before. I don’t expect you to have read them all but, as I said at the end of last year, it is a fag to have to state over and over something that has been said before. Hence my need to distill what I have written into a useful summary.

      On the whole I expect readers to do their own homework. I expect them also to take up with the originators of data or argument the errors or problems that they see in such data or argument. You will find Spencer and Christy, in this case, to be helpful and polite to a request for elaboration in the case you are putting forward.

      • bobo says:

        Thanks for your reply Don. One way to present your statement could be simply as a blog post with a link button to it on the menu bar at the top, with a comments section for people who want to quibble.

        How do you square up your claim (i) that data isn’t accurate or precise enough with your claim (iv) that you believe some warming has occurred over the 20th century? Can you see there seems to be a contradiction there? Also, if the data isn’t accurate enough why is a multidecadal positive trend emerging in so many independently collected data sets?

        Regarding UAH data, it appears you choose to use Spencer’s unverified data not the officially released data. The raw data for these sets is identical yet the two sets are very different. I’d be interested to know why you prefer to reference Spencer’s unverified set over the official set. BTW I’ll attempt to post a comment on his website regarding the odd data being officially published by UAH in reports he writes.

        Also, I still don’t understand why you have quoted BOM data when you apparently have some scepticism about it.

        • donaitkin says:

          Thanks for the suggestion. I will discuss this with my web-designer in due course.
          Why do I accept that there has been some warming? Well, there is abundant evidence that there has been a melting of glaciers in different parts of the world, not all of them and not all of the time. The general trend of all datasets is upwards in temperature, not all of them all of the time. That would do for me.

          That does not commit me to agreeing that any of the datasets is right at any time, or that x year is the most hottest ever.

          And all that is without relevance to more interesting questions, like whether or not warming is better, all things considered, and that higher CO2 levels might also be better for eco-systems.

          But it does bring into question the amazing confidence of some politicians and some scientists that they are quite sure about all of this. I find it hard to see how any rational person can be so sure, given the data.

          And it brings into question the supposed utility of carbon taxes, ETS and their ilk.

          • bobo says:

            Don, in that case, you probably should clarify what you mean by

            (i) the data simply do not allow accurate, precise accounts of global temperature, for all sorts of reasons

            What results are being extracted from data that you think are unjustified because, in your view, the data uncertainties are too large to make such claims? How much smaller would the uncertainty need to be before such results were acceptable?

  • bobo says:

    Don, much of blog post is a straw man fallacy by the way. You devote a great deal of effort to ridiculing the idea that the recent bad weather in various places is caused by global warming. But can you reference a study or a climate scientist who has specifically said that this is the case?

    The claims about extreme weather events in climate science are essentially restricted to talking about long term trends, and climate science literature is clear about that.

  • NameGlenM says:

    What does it matter if temperatures have risen-or fallen over whatever time-scales.Attribution remains the issue but I lean towards other mechanisms other than mans signature apropos CO2.

  • NameGlenM says:

    I’ll hedge David except to say it’s most likely(not entirely) oceanic turnover.

    • bobo says:


      Thermal energy is accumulating in the oceans, the oceans are certainly not losing net energy to the atmosphere/cryosphere/land

      • NameGlenM says:

        Thermal energy has always been.How it ends up in deep ocean is questionable.Anyway it’s business as usual for the planet- mostly staying within( ahem) acceptable parameters.If anyone can put up hard evidence that anthropogenic CO2 is or will cause catastrophic climate- change.Start with CO2 and temperature.It has to be CO2 because it can’t be anything else is not an excuse.

  • donaitkin says:

    Bobo,about SST.

    From what position of advantage are you speaking? Do you know anything, have you read anything, about the difficulties of coverage and the varieties of technique? I have.

    Stop trying to sound superior, and go and read in the literature.

    Start with ftp://ftp.wmo.int/Documents/PublicWeb/amp/mmop/documents/JCOMM-TR/J-TR-13-Marine-Climatology/REV1/joc1169.pdf

    and the various sites, like ICOADS where you can find out more about sea-surface temperatures (which, incidentally, are never exactly that).

    If you disagree, put forward your own views, with supporting evidence. I’m sure readers would be interested to see them.

    • bobo says:

      Apologies Don if I came across as obnoxious in my comment.

      I’m just not sure what point you’re trying to make about the ICOADS SST data. Of course there is uncertainty, more as you go back in the records, but I’m unclear how the uncertainty is so large that all sorts of conclusions have no basis.

      It seems you have some vaguely stated issue with the interpolating method to fill the gaps between the discrete measurement points and the uncertainties associated with these as well as with the temperature measurement uncertainties. Monte Carlo simulations can be used to generate temperature functions subject to various uncertainty and interpolating model constraints, and these are used to deduce global uncertainty values. That’s all I can really say because I’m not sure the precise nature of your objection. There is a highly cited paper which goes into this in detail:


      If you can be more precise with the nature of your doubt in this matter I will be happy to do a bit of reading on this.

      • bobo says:

        Oops the paper isn’t “highly cited”, there 6 citations in Google Scholar. I don’t have Web of Science in front of me.

      • donaitkin says:

        Bobo, I was simply asking for some common sense. The Southern Hemisphere is mainly ocean and its oceans much less the less sailed over. Until recently it had fewer data points.

        If you think that it is possible to construct persuasive estimates even of temperature given all the other variables, let alone the error bars, I will give you the website to set out your views, and how you would overcome all the difficulties that are there.

        I know that people have made estimates of temperature, though not of error bars, at least published ones. I find them quite unpersuasive. If you disagree, you might tell us why.

        • bobo says:

          Don, I’ll look at a few papers/chapters and post my thoughts at some point, there’s plenty of stuff I don’t know about these reconstructions so it could take a bit of time.

          What are your thoughts on the Berkeley Earth analysis and the consistency of this with other independently treated records such as NOAA, HadCRUT etc?

  • David says:

    Don In response to your comment above


    I am not here to engage in a quiz show…..

    But each time you do that you are introducing measurement error….


    This is where we disagree. Simply telling me there are measurement errors does not constitute an argument. If the errors are random they will sum to zero and the underlying trend will reveal itself. The statistical analysis in this case is fairly straight forward.

    My default position is to assume error are random, while accepting that this might not be correct. But if you think there is some systematic error that introduces bias to temperature trend (e.g. like Judith Curry), that is another argument all together. You may be right. But you will need to construct an argument along those lines. You don’t.

    For example Mann argued that Dr Spencer did not take into account that the satellite were slowly loosing height hence the data reported a spurious cooling trend. This is an argument based on bias.

    Or Professor Curry argues that “other” factors drive temperature increases. This is also an argument about bias.

    Either way I know of no valid statistical test that compares an “error bar” with an “increase” to test for bias or error the way you propose.

Leave a Reply