http://newsvote.bbc.co.uk/2/hi/scien...re/7786060.stm

Just looking at the figures on the above link.....

Can anybody explain how and why two significant decimal places are
considered? As far as I recollect an ordinary mercury thermometer is
accurate to 0.2 deg. So to suggest global mean temperatures are 14.31
degs seems totally meaningless. I would have though just 14 degs is
enough without trying to invent a greater accuracy.

On Wed, 17 Dec 2008 07:39:37 -0800 (PST), Pete L wrote in


http://newsvote.bbc.co.uk/2/hi/scien...re/7786060.stm

Just looking at the figures on the above link.....

Can anybody explain how and why two significant decimal places are
considered? As far as I recollect an ordinary mercury thermometer is
accurate to 0.2 deg. So to suggest global mean temperatures are 14.31
degs seems totally meaningless. I would have though just 14 degs is
enough without trying to invent a greater accuracy.

I think that precision is justified when it is based on hundreds or
thousands of mean temperatures.

--
Mike Tullett - Coleraine 55.13°N 6.69°W posted 17/12/2008 17:19:42 GMT

On 17 Dec, 17:19, Mike Tullett wrote:
On Wed, 17 Dec 2008 07:39:37 -0800 (PST), Pete L wrote in


http://newsvote.bbc.co.uk/2/hi/scien...re/7786060.stm

Just looking at the figures on the above link.....

Can anybody explain how and why two significant decimal places are
considered? As far as I recollect an ordinary mercury thermometer is
accurate to 0.2 deg. So to suggest global mean temperatures are 14.31
degs seems totally meaningless. I would have though just 14 degs is
enough without trying to invent a greater accuracy.

I think that precision is justified when it is based on hundreds or
thousands of mean temperatures.

--
Mike Tullett - Coleraine 55.13°N 6.69°W *posted 17/12/2008 17:19:42 *GMT

So, you suggest that the error cancels itself out with thousands of
mean temps? That supposes that there is an equal number of
thermometers reading high and low. Seems a bit unlikely.

"Pete L" wrote:
Just looking at the figures on the above link.....

Can anybody explain how and why two significant decimal places are
considered? As far as I recollect an ordinary mercury thermometer is
accurate to 0.2 deg. So to suggest global mean temperatures are 14.31
degs seems totally meaningless. I would have though just 14 degs is
enough without trying to invent a greater accuracy.

Pete, the fact is that the 14.31 is not a real temperature. It
involves not only temperature, but also time and space. It is,
perhaps, best to regard it as an index of planetary warmth or
coldness over a given period. Some might even argue that it serves
only to confuse that it looks like a temperature, and that giving it
a unit (i.e. °C) adds to the confusion. The accuracy of individual
thermometers and the precision to which they may be read
are neither here nor there.

Philip

On 17 Dec, 18:54, Pete L wrote:
On 17 Dec, 17:19, Mike Tullett wrote:





On Wed, 17 Dec 2008 07:39:37 -0800 (PST), Pete L wrote in


http://newsvote.bbc.co.uk/2/hi/scien...re/7786060.stm

Just looking at the figures on the above link.....

Can anybody explain how and why two significant decimal places are
considered? As far as I recollect an ordinary mercury thermometer is
accurate to 0.2 deg. So to suggest global mean temperatures are 14.31
degs seems totally meaningless. I would have though just 14 degs is
enough without trying to invent a greater accuracy.

I think that precision is justified when it is based on hundreds or
thousands of mean temperatures.

--
Mike Tullett - Coleraine 55.13°N 6.69°W *posted 17/12/2008 17:19:42 *GMT

So, you suggest that the error cancels itself out with thousands of
mean temps? That supposes that there is an equal number of
thermometers reading high and low. Seems a bit unlikely.- Hide quoted text -

- Show quoted text -

It does minimise the error. I read my barometer to the nearest mb. If
I had 1,000 barometers calibrated to the same standard, and managed to
read them all at the same time, and took the mean, it would be
reasonable to quote it to a higher level of precision.

The more readings, the more meaningful the result. So, if the MetO had
sites in St. Mawgan, Falmouth, St Ives, Penzance . . . . . - Sorry, a
brief drift into fantasy.

Graham
Penzance

If standard error of station mean temperature is say 0.1 degC, then
divide by square root of no. of stations, so 100 stations give 0.01
degC. Probably not worth giving more accuracy because of possible
bias due to changes in thermometer calibration and observing
practices, and locations of stations. I hope that the climate
scientists have considered this, and that any bias over the years is
~0.01 rather than ~0.1 degC.
Keith Grant

wrote in message
...
If standard error of station mean temperature is say 0.1 degC, then
divide by square root of no. of stations, so 100 stations give 0.01
degC. Probably not worth giving more accuracy because of possible
bias due to changes in thermometer calibration and observing
practices, and locations of stations. I hope that the climate
scientists have considered this, and that any bias over the years is
~0.01 rather than ~0.1 degC.
Keith Grant
-------------------
Yes and Philip is surely right in that the number of readings, if as high as
stated, gives the "value" high precision but as there is no "true mean" it's
accuracy cannot be quantified.
Dave

On Dec 17, 7:06*pm, "Philip Eden" philipATweatherHYPHENukDOTcom
wrote:
"Pete L" wrote:
Just looking at the figures on the above link.....

Can anybody explain how and why two significant decimal places are
considered? As far as I recollect an ordinary mercury thermometer is
accurate to 0.2 deg. So to suggest global mean temperatures are 14.31
degs seems totally meaningless. I would have though just 14 degs is
enough without trying to invent a greater accuracy.

Pete, the fact is that the 14.31 is not a real temperature. It
involves not only temperature, but also time and space. It is,
perhaps, best to regard it as an index of planetary warmth or
coldness over a given period. Some might even argue that it serves
only to confuse that it looks like a temperature, and that giving it
a unit (i.e. °C) adds to the confusion. The accuracy of individual
thermometers and the precision to which they may be read
are neither here nor there.

Philip

I really don't quite get this. The 14.31, whatever its accuracy,
is as real a temperature as any other mean, eg the mean temperature in
Nether Wallop in May 1955. There is no other way of expressing it.
The accuracy quoted is genuine only if both the errors of the
thermometers and the errors of reading them are symmetrically
distributed about zero. If there is a systematic bias in the
thermometers the final average will be in error by the same amount
whatever the number of readings. On the other hand one could argue
that if there has been a systematic bias in the either the
thermometers or the observers' readings and over the years neither of
these had changed then the quoted figure is OK because its sole
purpose is one of camparison over time. In any case the inaccuracies
introduced by inadequate sampling over the globe must exceed any of
the above errors.

Tudor Hughes, Warlingham, Surrey.

On Wed, 17 Dec 2008 at 12:08:18, Graham Easterling
wrote in uk.sci.weather :

It does minimise the error. I read my barometer to the nearest mb. If
I had 1,000 barometers calibrated to the same standard, and managed to
read them all at the same time, and took the mean, it would be
reasonable to quote it to a higher level of precision.

I was always told that you mustn't quote an average to a greater
precision than the accuracy of the instrument. That doesn't necessarily
mean I abide by that, though...
--
Paul Hyett, Cheltenham (change 'invalid83261' to 'blueyonder' to email me)

"Paul Hyett" wrote in message
...
On Wed, 17 Dec 2008 at 12:08:18, Graham Easterling
wrote in uk.sci.weather :

It does minimise the error. I read my barometer to the nearest mb. If
I had 1,000 barometers calibrated to the same standard, and managed to
read them all at the same time, and took the mean, it would be
reasonable to quote it to a higher level of precision.


But that is not a real world situation. In reality the standard errors would
be different, hopefully they would come from the same population but they
will still be different for a variety of reasons.

I was always told that you mustn't quote an average to a greater precision
than the accuracy of the instrument.

That's true Paul.
Keith Grant, who posted earlier in the thread is the expert and he is
absolutely correct.

Will
--