In article ,
Darren Prescott writes:
snip
Beyond that the operational GFS
restores an easterly flow, although that isn't supported by the ensembles
which suggest a somewhat more mobile outlook.

I think that the GFS ensemble consists of 10 runs, doesn't it (including
the operational and the "control" runs)? Assuming for the sake of
argument that I'm right, then I wonder how often the operational run is
the "best" of the ten. It should be at least one in ten, but it ought to
be a bit higher than that since its "unperturbed" data is theoretically
the best. But how much higher than 10%? Would it perform the best on 11%
of occasions? On 15%? On 20%? I'd be surprised if it was much more than
20%. I imagine that those responsible for the GFS must collect
statistics on this, but I don't suppose that they are made public.

Of course, if the operational run shows one thing, and all the other
nine runs are united in showing something else, then the chance that the
operational run is right is likely to be considerably less than 10%. But
that should be rare.
--
John Hall

"I am not young enough to know everything."
Oscar Wilde (1854-1900)