Why 1961-1990?
Graham Easterling
( www.metoffice.gov.uk/climate/uk/2008/september.html as an example )
As I understand it the 30 year period was brought into use as it is
long enough to produce an average which can be considered 'normal',
without being unduly outdated by climatic change. Surely the most
recent 30 years should be used?
In fact, bearing in mind the difference between the 60's & 70's
compared with more recent years, there is an argument for using a
slightly shorter period (Say 1980-2005)
Graham
Penzance
cumu...@yahoo.com.au
> Is there a reason why the MetO still use 1961-90 for 'normal' weather?
>
> As I understand it the 30 year period was brought into use as it is
> long enough to produce an average which can be considered 'normal',
> without being unduly outdated by climatic change. Surely the most
> recent 30 years should be used?
>
> In fact, bearing in mind the difference between the 60's & 70's
> compared with more recent years, there is an argument for using a
> slightly shorter period (Say 1980-2005)
>
> Graham
> Penzance
Graham, WMO defined various 'standard' 30 year periods over 40 years
ago, the first being 1931-60, the others defined as following on from
this but specified as being non-overlapping - hence 1961-90. According
to this arcane ruling, the next standard period averages will not be
updated until the period 1991-2020 is complete. (Even then no doubt
it'll be many years before the Met Office updates its averages.)
No doubt this made sense in the days when every national met service
had its own rules for averaging periods, and probably from an era when
everything had to be laboriously calculated by hand rather than in a
millisecond in a spreadsheet or database, but it certainly makes
little sense these days, particularly with rather rapid warming taking
place resulting in almost every month being well above the 1961-90
normal for years on end.
WMO really should look at changing the recommendation to 'the most
recent 30 year period ending in a decade' (which now would mandate
1971-2000 of course, which is now pretty much standard anyway - but
the Met Office still uses a bizarre mix of 1961-90 and 1971-2000
averages, for no very obvious reason). This would at least mandate a
reasonably topical averaging period. At least the next averaging
period under this method isn't very far away - 1981-2010 is fast
approaching.
The reason it's a 30 year period comes from statistical theory - from
a theoretically infinitely large dataset with a stable mean (dubious
with temperatures now of course), the mean of 30 samples should give
an average practically indistinguishable from the true population
mean. Victorian meteorologists worked this out empirically (although
they used 35 years - 1881-1915 being one such period), and it was
later proved on theoretical grounds that 30 was sufficient. And the
reason it ends in a decade is just tidiness, I guess ... !
HTH.
--
Stephen Burt
Stratfield Mortimer, Berkshire
Jack (jack.harrison@gmail.com)
say the past 30 years, ie during October 2008, the rolling period
would be October 1979 to September 2008 inclusive. In November 2008,
October 1979 would have been removed from the calculation and then the
dates November 1979 to September 2008 inclusive used.
An even simpler year rolling period could be used, ie during 2008 the
years 1978 to 2007 inclusive.
Jack
Jack (jack.harrison@gmail.com)
use period 1978-2007. Then next year, use 1979-2008. And so on.
Jack
John Dann
informative to use two separate 30-year reference periods, say 1961-90
as an absolute (ie unchanging) benchmark period and the most recent
rolling 30-year period as a relatively recent 'current context'
period. Of course you could debate indefinitely whether 1961-90 ought
to be the constant reference rather than some other tridecadal (if
that's indeed a word known to the English language) period, but being
as 1961-90 is the period that seems to be in regular use right now
then why not stick with it.
JGD
Weatherlawyer
> On 3 Oct, 20:32, Graham Easterling <GEasterl...@googlemail.com> wrote:
>
> > Is there a reason why the MetO still use 1961-90 for 'normal' weather?
>
> > As I understand it the 30 year period was brought into use as it is
> > long enough to produce an average which can be considered 'normal',
> > without being unduly outdated by climatic change. Surely the most
> > recent 30 years should be used?
>
> > In fact, bearing in mind the difference between the 60's & 70's
> > compared with more recent years, there is an argument for using a
> > slightly shorter period (Say 1980-2005)
>
> ago, the first being 1931-60, the others defined as following on from
> this but specified as being non-overlapping - hence 1961-90. According
> to this arcane ruling, the next standard period averages will not be
> updated until the period 1991-2020 is complete. (Even then no doubt
> it'll be many years before the Met Office updates its averages.)
I don't see what is so arcane about establishing work-bases.
If you just use random dates, the danger of finding ones to suit is
pernicious. Having data sets at least stops the range of statistics
overwhelming.
One might even be able to picture event sequences.
Unlikely but possible.
> No doubt this made sense in the days when every national met service
> had its own rules for averaging periods, and probably from an era when
> everything had to be laboriously calculated by hand rather than in a
> millisecond in a spreadsheet or database, but it certainly makes
> little sense these days, particularly with rather rapid warming taking
> place resulting in almost every month being well above the 1961-90
> normal for years on end.
>
> WMO really should look at changing the recommendation to 'the most
> recent 30 year period ending in a decade' (which now would mandate
> 1971-2000 of course, which is now pretty much standard anyway - but
> the Met Office still uses a bizarre mix of 1961-90 and 1971-2000
> averages, for no very obvious reason). This would at least mandate a
> reasonably topical averaging period. At least the next averaging
> period under this method isn't very far away - 1981-2010 is fast
> approaching.
>
> The reason it's a 30 year period comes from statistical theory - from
> a theoretically infinitely large dataset with a stable mean (dubious
> with temperatures now of course), the mean of 30 samples should give
> an average practically indistinguishable from the true population
> mean. Victorian meteorologists worked this out empirically (although
> they used 35 years - 1881-1915 being one such period), and it was
> later proved on theoretical grounds that 30 was sufficient. And the
> reason it ends in a decade is just tidiness, I guess ... !
Contemplating one's navel is a sign of peace and plenty. One needs
plenty of food and beer and no aggro in order to do it well.
Contemplating one's anus merely requires something to base statistics
on. And the urge to lose ones' self in a singularity.
Having said that, the 30 year period suits lunar cycles quite well.
Weatherlawyer
But there is a roll to it. 1990 to 2008 is oer half the cycle and can
be used by sadsadisticans.
Hell, the buggers will use anything, they are not constrained by
meaning.
Or usefulnessnesses.
Graham Easterling
> Of course you could debate indefinitely whether 1961-90 ought
> to be the constant reference rather than some other tridecadal (if
> that's indeed a word known to the English language) period, but being
> as 1961-90 is the period that seems to be in regular use right now
> then why not stick with it.
>
Because it is almost invariably associated with the word 'normal', and
what was normal for 1961-90 certainly (as far as temperatures are
concerned) has not been the normal for the past 20 years.
Giving Penzance as an example ( see www.easterling.freeserve.co.uk/wpage2.html
- bottom of page. )
A summary of average annual temperatures being:-
1861-1880 11.1
1961-1990 11.1 (surprisingly no change)
but
1991-2007 11.9
Graham
Penzance
cumu...@yahoo.com.au
Actually - this *was* done a long time ago, when the 1881-1915 period
was retained as the 'standard' averaging period, followed by 1916-50.
Averages for the earlier period were still in use up to the early
1970s, remarkably, while 1916-50 can still be found in some
publications today, particularly for rainfall averages. However, one
problem is that stations open and close, and after a while the number
of sites open for both the reference period and for more recent
periods becomes very small, making comparisons difficult. Sooner or
later the 'reference' period has to be updated - better to have it a
recent 30 year period, changed every 10 years, than something that we
know no longer represents current climat, IMHO.