Click images for more details



Recent comments
Recent posts
Currently discussing

A few sites I've stumbled across recently....

Powered by Squarespace

Discussion > I'll bet not a lot of people know this

Martin A

You estimate that 60 thermometers is too small a sample to calculate a global average.

How many do you estimate are necessary? Please show your calculation.

Apr 14, 2016 at 5:50 PM | Unregistered CommenterEntropic man

I'll get you started with the locations, EM.

With the area of the USA being about equal to about 2% of the worlds area, that might give them 1 or 2, max, of Phil Clarke and aTTP's 60 thermometers.

It seems apparent they would place one those thermometers in the Oval Office, and the other one in Harry Nyquist's grave. How on earth they could expect to get a meaningful measurement of US temperatures from these two thermometers beggars belief. I mean, what about all the thermometers that need to placed in Michael Mann's and Gavin Schmidt's fundaments?

More politely, but no less seriously, I thought even you had recently started to realise the difficulties of calculating a "global temperature" from insufficient data points. Any scientist or engineer who has measured something in the real world would understand this. What it says about those two isn't repeatable here. You don't need a detailed understanding of information theory when something doesn't even pass the smell-test.

Apr 14, 2016 at 7:15 PM | Unregistered Commentermichael hart


I thought even you had recently started to realise the difficulties of calculating a "global temperature"

I thought that even you had recently started to realise that we're not really calculating a "global temperature".

Apr 14, 2016 at 7:23 PM | Unregistered Commenter...and Then There's Physics

I thought that even you had recently started to realise that we're not really calculating a "global temperature".

aTTP - Could you elaborate? I had thought that was what it was all about (climate sensitivity, the Met Office saying it was the highest ever, etc).

Maybe I have missed a point. What is it that we are now really calculating if not 'global temperature' ? I certainly have not started to realise something different is now being calculated.

Apr 14, 2016 at 7:41 PM | Registered CommenterMartin A

And you didn’t think my response was also meant to be funny?! Dang!

RR - I don't think I said that I didn't think that. (If you follow me)

Apr 14, 2016 at 7:43 PM | Registered CommenterMartin A

aTTP. But the public are being told that you guys are indeed calculating a global temperature, and that it is increasing. Politicians react to these measurements and propose public policy changes that have already had adverse effects on industry and economies. Now you tell us we can yet useful information from only 60 thermometers, but we are not seeking a global temperature.

Something smells. Either that or everything I have read about AGW is wrong.

Apr 14, 2016 at 7:47 PM | Unregistered CommenterAlan Kendall

But it's not a "global temperature", it's a "global temperature anomaly". The key point is this. If you really wanted a global temperature, you would indeed probably need a lot of measurements and need to define precisely what it is you were measuring (temperature at 6am, 2m above the ground, at a location a certain distance from any buildings or structures,....). However, if you simply want to know if it is changing, then that isn't necessary. As long as you have records from which you can determine if it's gone up, down, or stayed the same, then that should be sufficient. Also, as Phil Clarke shows, there is work that has indicated that changes in temperature are correlated over distances of about 1200km, or smaller. Therefore you don't even need very many stations to produce a dataset that can tell you if global temperatures are changing. Of course, the more stations you have, the better it will probably be, but there are studies showing that you really only need about 60.

Apr 14, 2016 at 7:52 PM | Unregistered Commenter...and Then There's Physics

Wow, my flippant post has shown that ATTP, EM and Co have no panties.

Myself, I am glad that DrC and Martin have not called me out

Apr 14, 2016 at 8:24 PM | Unregistered CommenterDiogenes

Hat tip to Micky H Corbett for reminding me of the Central limits theorem as applied to determining a required sample size.. It allows an approximate back-of-the-envelope calculation of the required sample size (n) for a global mean temperature determination.

First some parameters.

If you use a mercury thermometer you measure to the nearest degree. Call the standard deviation (sigma) of your measurements 0.25C.

For the 95% confidence limit of the mean (W) I shall use the same limits quoted for GISS, ie +/-0.09C.

From the central limits theorem




In this case n=16*0.25^2/0.09^2 = 123

That is twice the estimate of 60 stations from Phil Clarke's link but still way below the sort of numbers Martin A's incredulity would presumably require.(I note that Martin A does not actually specify a number)

If you go to GISS you will find that their global temperatures are based on about 2200 stations, well above the minimum required for a reliable global temperature mean or global temperature anomaly .

Apr 14, 2016 at 8:45 PM | Unregistered CommenterEntropic man


I'm just Micky. I don't typically use the Dr unless I need to.

Alan I pointed this out earlier but their particular method especially using anomalies is the CLT. They just don't have large uncertainties. When you add them in then it would be more in realistic.

Apr 14, 2016 at 8:52 PM | Registered CommenterMicky H Corbett


The numbers show that you are the one with the bare arse.

Apr 14, 2016 at 8:59 PM | Unregistered CommenterEntropic man

aTTP. O.K. so you're investigating temperature anomalies now, because it's not possible to adequately determine a global temperature, but a global change is possible. Good luck selling that. You can't determine temperature , yet that is what has been sold to Joe Public. I think you have a PR disaster awaiting you.

Then you accept a study that implies a temperature anomaly can represent a region up to 1200km away from a single measurement point. Yet every year the UK Met Office puts out maps showing large variations in average temperatures across the relatively small area of the UK (with perhaps Scotland being warmer than usual, whereas the South East was cooler). Yet you are suggesting that a single measuring point, say in the Midlands, could represent this diversity. Again I am incredulous.

It's all getting to sound like madness, and it will take a lot to convince me. But then I have to accept quantum mechanics and that is truly mad.

Apr 14, 2016 at 10:47 PM | Unregistered CommenterAlan Kendall

Are you really saying that you spent years working at UEA and didn't know that when people talk about changes to global temperatures, they mean anomalies? Did you also not talk to your colleagues there? Do you also think that if we measure a change in global temperature that it hasn't actually changed? There are perfectly valid reasons for using anomalies, rather than trying to deal with absolute temperatures.

Apr 14, 2016 at 11:10 PM | Unregistered Commenter...and Then There's Physics

So anomalies have no bearing on temperatures?

Apr 15, 2016 at 1:14 AM | Unregistered Commenterdiogenes

aTTP. There are valid reasons why I do not wish to answer the personal questions (apart from their relevance) but why didn't you answer mine?

It seems to be intuitive that if one cannot establish a global average temperature accurately, then it will be impossible to obtain other such measurements at different times, and that you need these in order to obtain a temperature change.(= anomaly). Furthermore since each temperature estimation has a error associated with it the error for the anomaly should be greater. I agree that detecting an anomaly and its sign might be easy, but not its magnitude.

I note no response re the ability of a single measurement site to represent a large area.

Apr 15, 2016 at 6:34 AM | Unregistered CommenterAlan Kendall


aTTP. There are valid reasons why I do not wish to answer the personal questions (apart from their relevance) but why didn't you answer mine?

Okay, sorry, I wasn't meaning to ask what seemed to be personal questions.

It seems to be intuitive that if one cannot establish a global average temperature accurately, then it will be impossible to obtain other such measurements at different times, and that you need these in order to obtain a temperature change.

No, because if you have evidence that temperature changes are correlated over some distance (in this case, maybe 1200 km), then even a single station can tell you how temperatures have changed in a region (gone up, gone down, stayed the same) even if that single station would not necessarily tell you what the exact temperature would be in that region (since it is only one station). So, even reasonably sparse coverage can tell you if global temperatures have risen, fallen, or stayed the same, even if the coverage is insufficient to give you some kind of accurate absolute global temperature (rather than relative global temperature).

Apr 15, 2016 at 7:26 AM | Unregistered Commenter...and Then There's Physics


Ken is playing with you. He knows that the world of climate science talks of 'global temperature' and most people of an engineering or hard science background understand that the concept of 'one global temperature' is nonsense.

When you challenged him on this he says 'its anomalies silly!'.

Climate data aggregators use anomalies because output of different climate models vary so much that they resort to merging their anomaly output to achieve the 'correct' rate of rise in temperature.

As you pointed out the concept of using one measurement to represent the whole of the UK is a nonsense, but as Ken says, he has pointed to the evidence in a paper which states it is a valid technique, so it must be right!!!!!

The use of a single temperature measuring system covering the whole UK, is so wrong, whether you use anomalies or straight temperature reports.

Apr 15, 2016 at 10:39 AM | Unregistered CommenterSteve Richards

So when you see a global temperature graph, with a range that goes from -0.8 to 1.0 (ay GISSTEMP) do you really think that this means that the average global temperature is around 0? Or, do you think that maybe the data is relative to some baseline in the 20th century?

most people of an engineering or hard science background understand that the concept of 'one global temperature' is nonsense.

Is this some kind of appeal to some kind of authority? Let's bear in mind that many people with an engineering or hard science background to not regard the idea of a "global temperature" (as in anomalies) as nonsense. In what sense does your apparent background trump theirs?

Apr 15, 2016 at 10:53 AM | Unregistered Commenter...and Then There's Physics

Steve Richards, I suspected that my question put Ken in a quandary because he failed to address it twice. If he had tried I was going to counter that I don't even trust my single thermostat in the house to adequately control temperatures in all the different rooms. So why expect a single temperature reading location to adequately represent a large area. I really do question the common sense of some climatologists. It's as if they would accept that the temperature of a single house could represent an entire housing estate.

Thank you for your reply.

Apr 15, 2016 at 11:02 AM | Unregistered CommenterAlan Kendall

Actually Ken is quite correct in that you can represent say the temperature of the UK as a single value and furthermore as an anomaly. That is in line with averaging large data sets and removing part of them as constants.

The problem is that the uncertainty of that value will be so large compared to that needed to gain meaningful variation data from it that you may as well not produce the value.

Apr 15, 2016 at 11:14 AM | Registered CommenterMicky H Corbett

Mr Corbett: it does make sense that the fewer the data-points, the greater the uncertainty; while we certainly cannot claim accuracies to within ±0.1K, perhaps ±1K might be more acceptable. But then, that throws the whole thing into a quandary – how can they ramp up the fear of rises of 1.5K, ±1K? So, apart from the simple fact that they do not give the results that are so desperately sought, why do we not use vastly greater number of readings from the globally-encompassing satellites?

Martin A: you must understand that I understood that you understood that I understood that what you were saying was playing with what I was saying about playing with what you were saying. And if this horse twitches again, I’ll give you the whip.

Apr 15, 2016 at 11:27 AM | Registered CommenterRadical Rodent

Micky. Indeed it may be possible to produce a single average temperature (or anomaly) for the entire UK, computing this from numerous locations - but why do ir? The point is not whether this possible but whether a single temperature measuring location can represent the whole UK temperature with all its variety. The more I think about this the more nonsensical it seems.

Still no response from aTTP.

Apr 15, 2016 at 11:28 AM | Unregistered CommenterAlan Kendall

Still no response from aTTP.

This is rather typical blog comment behaviour. However, I thought I had responded, so you're going to have to explain what it is you would like me to respond to.

Apr 15, 2016 at 11:35 AM | Unregistered Commenter...and Then There's Physics

Look at it this way. Imagine you have 1000s of instruments measuring some quantity in a system. Lets imagine that each instrument measures to within some accuracy that is quite large with respect to the accuracy with which you want to measure some change in that system. Imagine at some instant in time you take a measurement with each instrument and you plot the distribution of the results. If you have enough measurements your distribution will probably be a normal distribution and you will be able to calculate the mean and a standard deviation. Now repeat this at a later time and imagine that the system has not changed at all. Each individual measurement might be different to the earlier one at the same location, but when you plot your distribution it will probably be the same as the earlier one (as long as you have enough measurements). So, you can again calculate the mean and standard deviation at this new time and you would expect them to be the same as before. If, however, the system has changed in some way (got warmer, colder,...) then you would expect a shift in the mean and you would expect to be able to determine a much smaller shift in the mean than the errors on the individual measurements, or than the standard deviation of the measurements at a single instant in time.

Apr 15, 2016 at 11:51 AM | Unregistered Commenter...and Then There's Physics

aTTP: a lot of imagination at play, there. Imagine, if you will, a uniform grid of observation stations over the surface of the Earth (the actual numbers a really irrelevant). Observations are being made at six-hourly intervals, to take into account diurnal variations; the instruments have all been verified and cross-referenced, to minimise instrument error, and the observers are trustworthy, and can be relied upon to make accurate measurements at the specified times. Imagine that there is no weather at all to influence the readings; the wind blows at a steady rate from the same direction all the time; cloud cover does not change, neither does precipitation. We can now start to construct a reasonably believable “global average” temperature to base further observations on; how long it will take to get this reference point is moot, but 30 years seems to be the commonly-accepted interval. Will the instruments maintain their level of accuracy over that time-period? If they are changed, how often are they changed? If they are changed, is there a transitory period where comparisons between the new and the old instrument(s) can be assessed? If moving the station becomes necessary, is there a transitory period when the readings from the old and the new sites can be compared? Then we can add in the unknown effects that weather might have on these readings…

Can you not see that there seems to be an awful lot at play? Yet you want to hold that this complex mish-mash is all down to greedy humans burning fossil fuels. So, the temperatures have risen a little since the little ice age? So, they might rise a bit more? As the rise so far has been on the whole beneficial, why the assumption that any further rise will lead to catastrophe? Given that we have only really started to study this phenomenon in the required detail only relatively recently (depending on your viewpoint or bias, anywhere from <30 years to >300 years), how can you be so convinced that it is down to one tiny proportion of the atmosphere; the proportion – surprise, surprise! – over which humans are perceived to have some control. That oh-so-convenient coincidence should set every sceptical and cynical alarm bell ringing in a semi-rational mind.

You like to point out the “planetary energy imbalance”, yet ignore my simple question on the previous page: when has there NOT been a planetary energy imbalance? That there have been temperature and climate changes since time immemorial, I would take a wild guess and answer that question with a simple: “Never!” What is there really to fear?

Apr 15, 2016 at 1:29 PM | Registered CommenterRadical Rodent

aTTP. A rather arrogant comment from you . (A rather typical blog behaviour) (Apr 15: 11.35am). This particularly so because 1) my comment was in my response to M.H.Corbett, which incorporated reference to the question not answered. and 2) I posed my question about the validity of a single measurement being able to represent the entire UK at 10.47pm and repeated my request for a response at 6.34 am. Only 5.01 hours later, you couldn't recall my question. It probably took you longer to type out your insulting 11.35am communication than scroll up to find my unanswered question.

This together with a previous response (11.10pm Apr 14) implying that I don't t know that climate science climate nsiders (inappropriately) that change = anomaly.

All in all not a very pleasant discussion.

Don't bother to reply now, Steve Richards was kind enough to confirm my suspicions.

Apr 15, 2016 at 2:18 PM | Unregistered CommenterAlan Kendall