On Sunday 31 Jan 2010 19:43, Len Wood scribbled:

Looking at model output every six hours in order to get a better idea
of what might happen beyond 5 or 6 days has been shown to be
pointless.
Doing this is verging on suffering from obsessive compulsive disorder.
The six hourly model runs are there to try to put a better feel of how
things might be up to five days ahead.
Beyond that once a day is quite sufficient. Slavishly looking every
six hours at what might happen in 10 days time shows a lack of
understanding of how these models are constructed

I think it can show the exact opposite. By looking at all runs instead of,
say, one run per day, you can get a better idea of the variation between the
runs and hence, perhaps, the reliability of a particular forecast - a bit
like a mental ensemble. A problem with looking at one run per day might be
that the run you pick is always a little biased.

A similar statement that I heard some time ago was regarding the frequency
that you should weigh yourself. It said that daily weighing was obsessive
and that once a month was ample. What this took no account of is daily
fluctuations in weight. If you weigh every day, you become aware of these
fluctuations and don't worry about them. If you only weigh yourself once a
month, you may get two extreme readings and worry about having put on a
couple of kilos. It's a bit like arguing that the climate has cooled since
1998. ;-)

--
Graham P Davis, Bracknell, Berks., UK. E-mail: newsman not newsboy
"I wear the cheese. It does not wear me."