The GFS isn't fit for purpose. Really, what is its value when it can
do a 180 degree turn before its even started. It's a waste of the USA
governments money and our time.

In article
,
Lawrence13 writes:
The GFS isn't fit for purpose. Really, what is its value when it can
do a 180 degree turn before its even started. It's a waste of the USA
governments money and our time.

It depends on what you think its purpose is. It almost always provides
useful guidance out to 5 days, usually to 7 or 8, and occasionally to 10
or more, especially from the ensemble as a whole. If you focus only on
the operational run, and expect reliable guidance towards the further
end of the 16 day period, then you are doomed to disappointment. But if
you think that the model is a waste of your time, then why spend time on
it? The remedy is in your own hands.

It's significant that the ECM doesn't release anything beyond 10 days,
and most of the other models restrict themselves to even shorter periods
than that. I'm sure that those responsible for the GFS regard the output
for beyond 10 days as a research tool rather than as suitable for making
a forecast, but being Americans they have an admirable attitude of
keeping nothing back from the general public. As to the cost to the
American government, I imagine that the extra expense of running the
model out to 16 days rather than 10 is minimal.

Whatever you think of Dawlish's attitude towards some other posters, his
requirement for consistency from four successive model runs before
taking what is shown seriously is a sensible one.
--
John Hall
"The power of accurate observation is commonly called cynicism
by those who have not got it."
George Bernard Shaw

John Hall wrote:
In article
,
Lawrence13 writes:
The GFS isn't fit for purpose. Really, what is its value when it can
do a 180 degree turn before its even started. It's a waste of the USA
governments money and our time.

It depends on what you think its purpose is. It almost always provides
useful guidance out to 5 days, usually to 7 or 8, and occasionally to 10
or more, especially from the ensemble as a whole. If you focus only on
the operational run, and expect reliable guidance towards the further
end of the 16 day period, then you are doomed to disappointment. But if
you think that the model is a waste of your time, then why spend time on
it? The remedy is in your own hands.

It's significant that the ECM doesn't release anything beyond 10 days,
and most of the other models restrict themselves to even shorter periods
than that. I'm sure that those responsible for the GFS regard the output
for beyond 10 days as a research tool rather than as suitable for making
a forecast, but being Americans they have an admirable attitude of
keeping nothing back from the general public. As to the cost to the
American government, I imagine that the extra expense of running the
model out to 16 days rather than 10 is minimal.

Whatever you think of Dawlish's attitude towards some other posters, his
requirement for consistency from four successive model runs before
taking what is shown seriously is a sensible one.
-----------------------------------------------------------
Ron and I went to Peter Gibb's talk at the RSC last night on forecasting
days ahead to decades ahead. He made it fairly clear that they had the
same problems with their model. He more or less said that it had
improved for forecasting up to 4-5 days ahead but after that confidence
was low. Sometimes in more predictable, stable types of weather it can
be extended but sometimes it can be even less. Of course there are the
ensembles and they would also look closely at the individual 50 members
to make a judgement on various potential outcomes. This is nothing new
to us of course but important nevertheless. They know the confidence (or
lack off ) in a particular forecast and he just wished they had more
than 1.5 minutes to deliver it as he would like to mention confidence
levels.
Dave