On Apr 7, 10:47*pm, "Norman" normanthis...@thisbitweather-
consultancy.com wrote:
Jon O'Rourke wrote:
"Dawlish" wrote in message

...

The operational run is the one that the computer
has selected as the most likely to achieve outcome

Is it ? I though it was essentially just another model run albeit at a
higher resolution compared to the ensemble, and even the operational
run lowers in resolution at T+180 IIRC.

Either way the GFS appears to be getting all the headlines in this
thread when the best model in terms of medium range deterministic and
ensemble output is ECMWF, IMHO of course but also according to
http://wwwt.emc.ncep.noaa.gov/gmb/ST.../aczhist6.html

Jon.

PS I thought 180C in Exeter was too warm last Thursday, roll on
autumn !

The GFS ensemble gets the headlines on here because the ECMWF ensemble
output isn't put into the public domain as far as I am aware.

Your PS strikes a chord!

Norman
--
Norman Lynagh
Chalfont St Giles, Buckinghamshire
85m a.s.l.
(remove "thisbit" twice to e-mail)- Hide quoted text -

- Show quoted text -

True, the gfs gets the headlines because it is the only model updated
4x per day - which is why I chose it for my forecasting. I'm well
aware nof the NOAA comparisons and my frustrations with being able to
find nothing, anywhere that goes beyond their 5-day-max horizon, just
added to my overall frustrations with the lack of accurate forecasting
at T240 (10 days). The ECM verifies, most of the time, as the best
forecasting model at 5 days, but even with the ECM, accuracy in
forecasting temperatures has varied between 0.91 and 0.72 over this
last month, with the mean accuracy being 0.847. Granted, this is
higher than the other 3 models that they evaluate (gfs in second place
with 0.793), but we are talking about the model being wrong about 15%
of the time even and only, at 5 days.

There are no published figures beyond this, though the fact that no-
one bothers to produce them, or comment upon them, has always
suggested to me that the accuracy falls off markedly between 5 days
and 10 days. The Met Office's lumping of all medium-term forecasts
into the 6-15 day range and the coining of the Internet terms; "FI"
and the model watcher's mantra; "beyond the reliable timeframe",
which is trotted out whenever someone disagrees with anything that
anyone else says at T144+, would support the assertion that accuracy
falls sharply. ECM may well be more accurate, but what level of
accuracy are we talking about? All models around 50%? Less?

So, how do we forecast with any believable accuracy at 10 days and how
do we measure it? Is it actually possible with present day forecasting
and model limitations?

Paul